US20190354628A1

US20190354628A1 - Asynchronous replication of synchronously replicated data

Info

Publication number: US20190354628A1
Application number: US16/050,698
Authority: US
Inventors: David Grunwald; Ronald Karr; Thomas Gill; Daquan Zuo
Original assignee: Pure Storage Inc
Current assignee: Pure Storage Inc
Priority date: 2018-05-21
Filing date: 2018-07-31
Publication date: 2019-11-21
Also published as: US11757795B2; US10992598B2; US20230396565A1; DE112019002584T5; WO2019226573A1; US20210409349A1; US20230344783A1; US20210243139A1; DE112019002609T5; US11677687B2; WO2019226586A1; US20190354450A1; US20190356609A1; US11128578B2; WO2019226597A1; CN112470142A

Abstract

A storage system asynchronously replicating a synchronously replicated dataset, where the asynchronous replication of the asynchronously replicated dataset includes: determining, at a target storage system, multiple work items corresponding to a dataset stored among multiple source storage systems, wherein each respective work item corresponds to a respective subset of the dataset; and for each session from among a plurality of sessions operating on the target storage system: determining one or more computing environment factors affecting performance of replication of data from one or more of the multiple source storage systems to the target storage system; identifying, for a given session and based on the one or more computing environment factors, a respective source storage system and a quantity of work items; and replicating, from the respective storage system, one or more subsets of data corresponding to the quantity of work items.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application is a non-provisional application for patent entitled to a filing date and claiming the benefit of earlier-filed U.S. Provisional Patent Application Ser. No. 62/674,570, filed May 21, 2018, and U.S. Provisional Patent Application Ser. No. 62/695,433, filed Jul. 9, 2018.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1A illustrates a first example system for data storage in accordance with some implementations.
FIG. 1B illustrates a second example system for data storage in accordance with some implementations.
FIG. 1C illustrates a third example system for data storage in accordance with some implementations.
FIG. 1D illustrates a fourth example system for data storage in accordance with some implementations.
FIG. 2A is a perspective view of a storage cluster with multiple storage nodes and internal storage coupled to each storage node to provide network attached storage, in accordance with some embodiments.
FIG. 2B is a block diagram showing an interconnect switch coupling multiple storage nodes in accordance with some embodiments.
FIG. 2C is a multiple level block diagram, showing contents of a storage node and contents of one of the non-volatile solid state storage units in accordance with some embodiments.
FIG. 2D shows a storage server environment, which uses embodiments of the storage nodes and storage units of some previous figures in accordance with some embodiments.
FIG. 2E is a blade hardware block diagram, showing a control plane, compute and storage planes, and authorities interacting with underlying physical resources, in accordance with some embodiments.
FIG. 2F depicts elasticity software layers in blades of a storage cluster, in accordance with some embodiments.
FIG. 2G depicts authorities and storage resources in blades of a storage cluster, in accordance with some embodiments.
FIG. 3A sets forth a diagram of a storage system that is coupled for data communications with a cloud services provider in accordance with some embodiments of the present disclosure.
FIG. 3B sets forth a diagram of a storage system in accordance with some embodiments of the present disclosure.
FIG. 4 sets forth a block diagram illustrating a plurality of storage systems that support a pod according to some embodiments of the present disclosure.
FIG. 5 sets forth a block diagram illustrating a plurality of storage systems that support a pod according to some embodiments of the present disclosure.
FIG. 6 sets forth a block diagram illustrating a plurality of storage systems that support a pod according to some embodiments of the present disclosure.
FIG. 7 sets forth diagrams of metadata representations that may be implemented as a structured collection of metadata objects that may represent a logical volume of storage data, or a portion of a logical volume, in accordance with some embodiments of the present disclosure.
FIG. 8 sets forth a diagram of a computing environment for asynchronously replicating a synchronously replicated dataset according to some embodiments of the present disclosure.
FIG. 9 sets forth a flow chart illustrating an example method of asynchronously replicating a synchronously replicated dataset according to some embodiments of the present disclosure.

DESCRIPTION OF EMBODIMENTS

Example methods, apparatus, and products for asynchronous replication of synchronously replicated data in accordance with embodiments of the present disclosure are described with reference to the accompanying drawings, beginning with FIG. 1A. FIG. 1A illustrates an example system for data storage, in accordance with some implementations. System 100 (also referred to as “storage system” herein) includes numerous elements for purposes of illustration rather than limitation. It may be noted that system 100 may include the same, more, or fewer elements configured in the same or different manner in other implementations.
System 100 includes a number of computing devices 164A-B. Computing devices (also referred to as “client devices” herein) may be embodied, for example, a server in a data center, a workstation, a personal computer, a notebook, or the like. Computing devices 164A-B may be coupled for data communications to one or more storage arrays 102A-B through a storage area network (‘SAN’) 158 or a local area network (‘LAN’) 160.
The SAN 158 may be implemented with a variety of data communications fabrics, devices, and protocols. For example, the fabrics for SAN 158 may include Fibre Channel, Ethernet, Infiniband, Serial Attached Small Computer System Interface (‘SAS’), or the like. Data communications protocols for use with SAN 158 may include Advanced Technology Attachment (‘ATA’), Fibre Channel Protocol, Small Computer System Interface (‘SCSI’), Internet Small Computer System Interface (‘iSCSI’), HyperSCSI, Non-Volatile Memory Express (‘NVMe’) over Fabrics, or the like. It may be noted that SAN 158 is provided for illustration, rather than limitation. Other data communication couplings may be implemented between computing devices 164A-B and storage arrays 102A-B.
The LAN 160 may also be implemented with a variety of fabrics, devices, and protocols. For example, the fabrics for LAN 160 may include Ethernet (802.3), wireless (802.11), or the like. Data communication protocols for use in LAN 160 may include Transmission Control Protocol (‘TCP’), User Datagram Protocol (‘UDP’), Internet Protocol (IF), HyperText Transfer Protocol (‘HTTP’), Wireless Access Protocol (‘WAP’), Handheld Device Transport Protocol (‘HDTP’), Session Initiation Protocol (‘SIP’), Real Time Protocol (‘RTP’), or the like.
Storage arrays 102A-B may provide persistent data storage for the computing devices 164A-B. Storage array 102A may be contained in a chassis (not shown), and storage array 102B may be contained in another chassis (not shown), in implementations. Storage array 102A and 102B may include one or more storage array controllers 110 (also referred to as “controller” herein). A storage array controller 110 may be embodied as a module of automated computing machinery comprising computer hardware, computer software, or a combination of computer hardware and software. In some implementations, the storage array controllers 110 may be configured to carry out various storage tasks. Storage tasks may include writing data received from the computing devices 164A-B to storage array 102A-B, erasing data from storage array 102A-B, retrieving data from storage array 102A-B and providing data to computing devices 164A-B, monitoring and reporting of disk utilization and performance, performing redundancy operations, such as Redundant Array of Independent Drives (‘RAID’) or RAID-like data redundancy operations, compressing data, encrypting data, and so forth.
Storage array controller 110 may be implemented in a variety of ways, including as a Field Programmable Gate Array (‘FPGA’), a Programmable Logic Chip (‘PLC’), an Application Specific Integrated Circuit (‘ASIC’), System-on-Chip (‘SOC’), or any computing device that includes discrete components such as a processing device, central processing unit, computer memory, or various adapters. Storage array controller 110 may include, for example, a data communications adapter configured to support communications via the SAN 158 or LAN 160. In some implementations, storage array controller 110 may be independently coupled to the LAN 160. In implementations, storage array controller 110 may include an I/O controller or the like that couples the storage array controller 110 for data communications, through a midplane (not shown), to a persistent storage resource 170A-B (also referred to as a “storage resource” herein). The persistent storage resource 170A-B main include any number of storage drives 171A-F (also referred to as “storage devices” herein) and any number of non-volatile Random Access Memory (‘NVRAM’) devices (not shown).
In some implementations, the NVRAM devices of a persistent storage resource 170A-B may be configured to receive, from the storage array controller 110, data to be stored in the storage drives 171A-F. In some examples, the data may originate from computing devices 164A-B. In some examples, writing data to the NVRAM device may be carried out more quickly than directly writing data to the storage drive 171A-F. In implementations, the storage array controller 110 may be configured to utilize the NVRAM devices as a quickly accessible buffer for data destined to be written to the storage drives 171A-F. Latency for write requests using NVRAM devices as a buffer may be improved relative to a system in which a storage array controller 110 writes data directly to the storage drives 171A-F. In some implementations, the NVRAM devices may be implemented with computer memory in the form of high bandwidth, low latency RAM. The NVRAM device is referred to as “non-volatile” because the NVRAM device may receive or include a unique power source that maintains the state of the RAM after main power loss to the NVRAM device. Such a power source may be a battery, one or more capacitors, or the like. In response to a power loss, the NVRAM device may be configured to write the contents of the RAM to a persistent storage, such as the storage drives 171A-F.
In implementations, storage drive 171A-F may refer to any device configured to record data persistently, where “persistently” or “persistent” refers as to a device's ability to maintain recorded data after loss of power. In some implementations, storage drive 171A-F may correspond to non-disk storage media. For example, the storage drive 171A-F may be one or more solid-state drives (‘SSDs’), flash memory based storage, any type of solid-state non-volatile memory, or any other type of non-mechanical storage device. In other implementations, storage drive 171A-F may include may include mechanical or spinning hard disk, such as hard-disk drives (‘HDD’).
In some implementations, the storage array controllers 110 may be configured for offloading device management responsibilities from storage drive 171A-F in storage array 102A-B. For example, storage array controllers 110 may manage control information that may describe the state of one or more memory blocks in the storage drives 171A-F. The control information may indicate, for example, that a particular memory block has failed and should no longer be written to, that a particular memory block contains boot code for a storage array controller 110, the number of program-erase (‘P/E’) cycles that have been performed on a particular memory block, the age of data stored in a particular memory block, the type of data that is stored in a particular memory block, and so forth. In some implementations, the control information may be stored with an associated memory block as metadata. In other implementations, the control information for the storage drives 171A-F may be stored in one or more particular memory blocks of the storage drives 171A-F that are selected by the storage array controller 110. The selected memory blocks may be tagged with an identifier indicating that the selected memory block contains control information. The identifier may be utilized by the storage array controllers 110 in conjunction with storage drives 171A-F to quickly identify the memory blocks that contain control information. For example, the storage controllers 110 may issue a command to locate memory blocks that contain control information. It may be noted that control information may be so large that parts of the control information may be stored in multiple locations, that the control information may be stored in multiple locations for purposes of redundancy, for example, or that the control information may otherwise be distributed across multiple memory blocks in the storage drive 171A-F.
In implementations, storage array controllers 110 may offload device management responsibilities from storage drives 171A-F of storage array 102A-B by retrieving, from the storage drives 171A-F, control information describing the state of one or more memory blocks in the storage drives 171A-F. Retrieving the control information from the storage drives 171A-F may be carried out, for example, by the storage array controller 110 querying the storage drives 171A-F for the location of control information for a particular storage drive 171A-F. The storage drives 171A-F may be configured to execute instructions that enable the storage drive 171A-F to identify the location of the control information. The instructions may be executed by a controller (not shown) associated with or otherwise located on the storage drive 171A-F and may cause the storage drive 171A-F to scan a portion of each memory block to identify the memory blocks that store control information for the storage drives 171A-F. The storage drives 171A-F may respond by sending a response message to the storage array controller 110 that includes the location of control information for the storage drive 171A-F. Responsive to receiving the response message, storage array controllers 110 may issue a request to read data stored at the address associated with the location of control information for the storage drives 171A-F.
In other implementations, the storage array controllers 110 may further offload device management responsibilities from storage drives 171A-F by performing, in response to receiving the control information, a storage drive management operation. A storage drive management operation may include, for example, an operation that is typically performed by the storage drive 171A-F (e.g., the controller (not shown) associated with a particular storage drive 171A-F). A storage drive management operation may include, for example, ensuring that data is not written to failed memory blocks within the storage drive 171A-F, ensuring that data is written to memory blocks within the storage drive 171A-F in such a way that adequate wear leveling is achieved, and so forth.
In implementations, storage array 102A-B may implement two or more storage array controllers 110. For example, storage array 102A may include storage array controllers 110A and storage array controllers 110B. At a given instance, a single storage array controller 110 (e.g., storage array controller 110A) of a storage system 100 may be designated with primary status (also referred to as “primary controller” herein), and other storage array controllers 110 (e.g., storage array controller 110A) may be designated with secondary status (also referred to as “secondary controller” herein). The primary controller may have particular rights, such as permission to alter data in persistent storage resource 170A-B (e.g., writing data to persistent storage resource 170A-B). At least some of the rights of the primary controller may supersede the rights of the secondary controller. For instance, the secondary controller may not have permission to alter data in persistent storage resource 170A-B when the primary controller has the right. The status of storage array controllers 110 may change. For example, storage array controller 110A may be designated with secondary status, and storage array controller 110B may be designated with primary status.
In some implementations, a primary controller, such as storage array controller 110A, may serve as the primary controller for one or more storage arrays 102A-B, and a second controller, such as storage array controller 110B, may serve as the secondary controller for the one or more storage arrays 102A-B. For example, storage array controller 110A may be the primary controller for storage array 102A and storage array 102B, and storage array controller 110B may be the secondary controller for storage array 102A and 102B. In some implementations, storage array controllers 110C and 110D (also referred to as “storage processing modules”) may neither have primary or secondary status. Storage array controllers 110C and 110D, implemented as storage processing modules, may act as a communication interface between the primary and secondary controllers (e.g., storage array controllers 110A and 110B, respectively) and storage array 102B. For example, storage array controller 110A of storage array 102A may send a write request, via SAN 158, to storage array 102B. The write request may be received by both storage array controllers 110C and 110D of storage array 102B. Storage array controllers 110C and 110D facilitate the communication, e.g., send the write request to the appropriate storage drive 171A-F. It may be noted that in some implementations storage processing modules may be used to increase the number of storage drives controlled by the primary and secondary controllers.
In implementations, storage array controllers 110 are communicatively coupled, via a midplane (not shown), to one or more storage drives 171A-F and to one or more NVRAM devices (not shown) that are included as part of a storage array 102A-B. The storage array controllers 110 may be coupled to the midplane via one or more data communication links and the midplane may be coupled to the storage drives 171A-F and the NVRAM devices via one or more data communications links. The data communications links described herein are collectively illustrated by data communications links 108A-D and may include a Peripheral Component Interconnect Express (‘PCIe’) bus, for example.
FIG. 1B illustrates an example system for data storage, in accordance with some implementations. Storage array controller 101 illustrated in FIG. 1B may similar to the storage array controllers 110 described with respect to FIG. 1A. In one example, storage array controller 101 may be similar to storage array controller 110A or storage array controller 110B. Storage array controller 101 includes numerous elements for purposes of illustration rather than limitation. It may be noted that storage array controller 101 may include the same, more, or fewer elements configured in the same or different manner in other implementations. It may be noted that elements of FIG. 1A may be included below to help illustrate features of storage array controller 101.
Storage array controller 101 may include one or more processing devices 104 and random access memory (‘RAM’) 111. Processing device 104 (or controller 101) represents one or more general-purpose processing devices such as a microprocessor, central processing unit, or the like. More particularly, the processing device 104 (or controller 101) may be a complex instruction set computing (‘CISC’) microprocessor, reduced instruction set computing (‘RISC’) microprocessor, very long instruction word (‘VLIW’) microprocessor, or a processor implementing other instruction sets or processors implementing a combination of instruction sets. The processing device 104 (or controller 101) may also be one or more special-purpose processing devices such as an application specific integrated circuit (‘ASIC’), a field programmable gate array (‘FPGA’), a digital signal processor (‘DSP’), network processor, or the like.
The processing device 104 may be connected to the RAM 111 via a data communications link 106, which may be embodied as a high speed memory bus such as a Double-Data Rate 4 (‘DDR4’) bus. Stored in RAM 111 is an operating system 112. In some implementations, instructions 113 are stored in RAM 111. Instructions 113 may include computer program instructions for performing operations in in a direct-mapped flash storage system. In one embodiment, a direct-mapped flash storage system is one that that addresses data blocks within flash drives directly and without an address translation performed by the storage controllers of the flash drives.
In implementations, storage array controller 101 includes one or more host bus adapters 103A-C that are coupled to the processing device 104 via a data communications link 105A-C. In implementations, host bus adapters 103A-C may be computer hardware that connects a host system (e.g., the storage array controller) to other network and storage arrays. In some examples, host bus adapters 103A-C may be a Fibre Channel adapter that enables the storage array controller 101 to connect to a SAN, an Ethernet adapter that enables the storage array controller 101 to connect to a LAN, or the like. Host bus adapters 103A-C may be coupled to the processing device 104 via a data communications link 105A-C such as, for example, a PCIe bus.
In implementations, storage array controller 101 may include a host bus adapter 114 that is coupled to an expander 115. The expander 115 may be used to attach a host system to a larger number of storage drives. The expander 115 may, for example, be a SAS expander utilized to enable the host bus adapter 114 to attach to storage drives in an implementation where the host bus adapter 114 is embodied as a SAS controller.
In implementations, storage array controller 101 may include a switch 116 coupled to the processing device 104 via a data communications link 109. The switch 116 may be a computer hardware device that can create multiple endpoints out of a single endpoint, thereby enabling multiple devices to share a single endpoint. The switch 116 may, for example, be a PCIe switch that is coupled to a PCIe bus (e.g., data communications link 109) and presents multiple PCIe connection points to the midplane.
In implementations, storage array controller 101 includes a data communications link 107 for coupling the storage array controller 101 to other storage array controllers. In some examples, data communications link 107 may be a QuickPath Interconnect (QPI) interconnect.
A traditional storage system that uses traditional flash drives may implement a process across the flash drives that are part of the traditional storage system. For example, a higher level process of the storage system may initiate and control a process across the flash drives. However, a flash drive of the traditional storage system may include its own storage controller that also performs the process. Thus, for the traditional storage system, a higher level process (e.g., initiated by the storage system) and a lower level process (e.g., initiated by a storage controller of the storage system) may both be performed.
To resolve various deficiencies of a traditional storage system, operations may be performed by higher level processes and not by the lower level processes. For example, the flash storage system may include flash drives that do not include storage controllers that provide the process. Thus, the operating system of the flash storage system itself may initiate and control the process. This may be accomplished by a direct-mapped flash storage system that addresses data blocks within the flash drives directly and without an address translation performed by the storage controllers of the flash drives.
The operating system of the flash storage system may identify and maintain a list of allocation units across multiple flash drives of the flash storage system. The allocation units may be entire erase blocks or multiple erase blocks. The operating system may maintain a map or address range that directly maps addresses to erase blocks of the flash drives of the flash storage system.
Direct mapping to the erase blocks of the flash drives may be used to rewrite data and erase data. For example, the operations may be performed on one or more allocation units that include a first data and a second data where the first data is to be retained and the second data is no longer being used by the flash storage system. The operating system may initiate the process to write the first data to new locations within other allocation units and erasing the second data and marking the allocation units as being available for use for subsequent data. Thus, the process may only be performed by the higher level operating system of the flash storage system without an additional lower level process being performed by controllers of the flash drives.
Advantages of the process being performed only by the operating system of the flash storage system include increased reliability of the flash drives of the flash storage system as unnecessary or redundant write operations are not being performed during the process. One possible point of novelty here is the concept of initiating and controlling the process at the operating system of the flash storage system. In addition, the process can be controlled by the operating system across multiple flash drives. This is contrast to the process being performed by a storage controller of a flash drive.
A storage system can consist of two storage array controllers that share a set of drives for failover purposes, or it could consist of a single storage array controller that provides a storage service that utilizes multiple drives, or it could consist of a distributed network of storage array controllers each with some number of drives or some amount of Flash storage where the storage array controllers in the network collaborate to provide a complete storage service and collaborate on various aspects of a storage service including storage allocation and garbage collection.
FIG. 1C illustrates a third example system 117 for data storage in accordance with some implementations. System 117 (also referred to as “storage system” herein) includes numerous elements for purposes of illustration rather than limitation. It may be noted that system 117 may include the same, more, or fewer elements configured in the same or different manner in other implementations.
In one embodiment, system 117 includes a dual Peripheral Component Interconnect (‘PCI’) flash storage device 118 with separately addressable fast write storage. System 117 may include a storage controller 119. In one embodiment, storage controller 119 may be a CPU, ASIC, FPGA, or any other circuitry that may implement control structures necessary according to the present disclosure. In one embodiment, system 117 includes flash memory devices (e.g., including flash memory devices 120 a-n), operatively coupled to various channels of the storage device controller 119. Flash memory devices 120 a-n, may be presented to the controller 119 as an addressable collection of Flash pages, erase blocks, and/or control elements sufficient to allow the storage device controller 119 to program and retrieve various aspects of the Flash. In one embodiment, storage device controller 119 may perform operations on flash memory devices 120A-N including storing and retrieving data content of pages, arranging and erasing any blocks, tracking statistics related to the use and reuse of Flash memory pages, erase blocks, and cells, tracking and predicting error codes and faults within the Flash memory, controlling voltage levels associated with programming and retrieving contents of Flash cells, etc.
In one embodiment, system 117 may include RAM 121 to store separately addressable fast-write data. In one embodiment, RAM 121 may be one or more separate discrete devices. In another embodiment, RAM 121 may be integrated into storage device controller 119 or multiple storage device controllers. The RAM 121 may be utilized for other purposes as well, such as temporary program memory for a processing device (e.g., a CPU) in the storage device controller 119.
In one embodiment, system 119 may include a stored energy device 122, such as a rechargeable battery or a capacitor. Stored energy device 122 may store energy sufficient to power the storage device controller 119, some amount of the RAM (e.g., RAM 121), and some amount of Flash memory (e.g., Flash memory 120 a-120 n) for sufficient time to write the contents of RAM to Flash memory. In one embodiment, storage device controller 119 may write the contents of RAM to Flash Memory if the storage device controller detects loss of external power.
In one embodiment, system 117 includes two data communications links 123 a, 123 b. In one embodiment, data communications links 123 a, 123 b may be PCI interfaces. In another embodiment, data communications links 123 a, 123 b may be based on other communications standards (e.g., HyperTransport, InfiniBand, etc.). Data communications links 123 a, 123 b may be based on non-volatile memory express (‘NVMe’) or NVMe over fabrics (‘NVMf’) specifications that allow external connection to the storage device controller 119 from other components in the storage system 117. It should be noted that data communications links may be interchangeably referred to herein as PCI buses for convenience.
System 117 may also include an external power source (not shown), which may be provided over one or both data communications links 123 a, 123 b, or which may be provided separately. An alternative embodiment includes a separate Flash memory (not shown) dedicated for use in storing the content of RAM 121. The storage device controller 119 may present a logical device over a PCI bus which may include an addressable fast-write logical device, or a distinct part of the logical address space of the storage device 118, which may be presented as PCI memory or as persistent storage. In one embodiment, operations to store into the device are directed into the RAM 121. On power failure, the storage device controller 119 may write stored content associated with the addressable fast-write logical storage to Flash memory (e.g., Flash memory 120 a-n) for long-term persistent storage.
In one embodiment, the logical device may include some presentation of some or all of the content of the Flash memory devices 120 a-n, where that presentation allows a storage system including a storage device 118 (e.g., storage system 117) to directly address Flash memory pages and directly reprogram erase blocks from storage system components that are external to the storage device through the PCI bus. The presentation may also allow one or more of the external components to control and retrieve other aspects of the Flash memory including some or all of: tracking statistics related to use and reuse of Flash memory pages, erase blocks, and cells across all the Flash memory devices; tracking and predicting error codes and faults within and across the Flash memory devices; controlling voltage levels associated with programming and retrieving contents of Flash cells; etc.
In one embodiment, the stored energy device 122 may be sufficient to ensure completion of in-progress operations to the Flash memory devices 107 a-120 n stored energy device 122 may power storage device controller 119 and associated Flash memory devices (e.g., 120 a-n) for those operations, as well as for the storing of fast-write RAM to Flash memory. Stored energy device 122 may be used to store accumulated statistics and other parameters kept and tracked by the Flash memory devices 120 a-n and/or the storage device controller 119. Separate capacitors or stored energy devices (such as smaller capacitors near or embedded within the Flash memory devices themselves) may be used for some or all of the operations described herein.
Various schemes may be used to track and optimize the life span of the stored energy component, such as adjusting voltage levels over time, partially discharging the storage energy device 122 to measure corresponding discharge characteristics, etc. If the available energy decreases over time, the effective available capacity of the addressable fast-write storage may be decreased to ensure that it can be written safely based on the currently available stored energy.
FIG. 1D illustrates a third example system 124 for data storage in accordance with some implementations. In one embodiment, system 124 includes storage controllers 125 a, 125 b. In one embodiment, storage controllers 125 a, 125 b are operatively coupled to Dual PCI storage devices 119 a, 119 b and 119 c, 119 d, respectively. Storage controllers 125 a, 125 b may be operatively coupled (e.g., via a storage network 130) to some number of host computers 127 a-n.
In one embodiment, two storage controllers (e.g., 125 a and 125 b) provide storage services, such as a SCS) block storage array, a file server, an object server, a database or data analytics service, etc. The storage controllers 125 a, 125 b may provide services through some number of network interfaces (e.g., 126 a-d) to host computers 127 a-n outside of the storage system 124. Storage controllers 125 a, 125 b may provide integrated services or an application entirely within the storage system 124, forming a converged storage and compute system. The storage controllers 125 a, 125 b may utilize the fast write memory within or across storage devices 119 a-d to journal in progress operations to ensure the operations are not lost on a power failure, storage controller removal, storage controller or storage system shutdown, or some fault of one or more software or hardware components within the storage system 124.
In one embodiment, controllers 125 a, 125 b operate as PCI masters to one or the other PCI buses 128 a, 128 b. In another embodiment, 128 a and 128 b may be based on other communications standards (e.g., HyperTransport, InfiniBand, etc.). Other storage system embodiments may operate storage controllers 125 a, 125 b as multi-masters for both PCI buses 128 a, 128 b. Alternately, a PCI/NVMe/NVMf switching infrastructure or fabric may connect multiple storage controllers. Some storage system embodiments may allow storage devices to communicate with each other directly rather than communicating only with storage controllers. In one embodiment, a storage device controller 119 a may be operable under direction from a storage controller 125 a to synthesize and transfer data to be stored into Flash memory devices from data that has been stored in RAM (e.g., RAM 121 of FIG. 1C). For example, a recalculated version of RAM content may be transferred after a storage controller has determined that an operation has fully committed across the storage system, or when fast-write memory on the device has reached a certain used capacity, or after a certain amount of time, to ensure improve safety of the data or to release addressable fast-write capacity for reuse. This mechanism may be used, for example, to avoid a second transfer over a bus (e.g., 128 a, 128 b) from the storage controllers 125 a, 125 b. In one embodiment, a recalculation may include compressing data, attaching indexing or other metadata, combining multiple data segments together, performing erasure code calculations, etc.
In one embodiment, under direction from a storage controller 125 a, 125 b, a storage device controller 119 a, 119 b may be operable to calculate and transfer data to other storage devices from data stored in RAM (e.g., RAM 121 of FIG. 1C) without involvement of the storage controllers 125 a, 125 b. This operation may be used to mirror data stored in one controller 125 a to another controller 125 b, or it could be used to offload compression, data aggregation, and/or erasure coding calculations and transfers to storage devices to reduce load on storage controllers or the storage controller interface 129 a, 129 b to the PCI bus 128 a, 128 b.
A storage device controller 119 may include mechanisms for implementing high availability primitives for use by other parts of a storage system external to the Dual PCI storage device 118. For example, reservation or exclusion primitives may be provided so that, in a storage system with two storage controllers providing a highly available storage service, one storage controller may prevent the other storage controller from accessing or continuing to access the storage device. This could be used, for example, in cases where one controller detects that the other controller is not functioning properly or where the interconnect between the two storage controllers may itself not be functioning properly.
In one embodiment, a storage system for use with Dual PCI direct mapped storage devices with separately addressable fast write storage includes systems that manage erase blocks or groups of erase blocks as allocation units for storing data on behalf of the storage service, or for storing metadata (e.g., indexes, logs, etc.) associated with the storage service, or for proper management of the storage system itself. Flash pages, which may be a few kilobytes in size, may be written as data arrives or as the storage system is to persist data for long intervals of time (e.g., above a defined threshold of time). To commit data more quickly, or to reduce the number of writes to the Flash memory devices, the storage controllers may first write data into the separately addressable fast write storage on one more storage devices.
In one embodiment, the storage controllers 125 a, 125 b may initiate the use of erase blocks within and across storage devices (e.g., 118) in accordance with an age and expected remaining lifespan of the storage devices, or based on other statistics. The storage controllers 125 a, 125 b may initiate garbage collection and data migration data between storage devices in accordance with pages that are no longer needed as well as to manage Flash page and erase block lifespans and to manage overall system performance.
In one embodiment, the storage system 124 may utilize mirroring and/or erasure coding schemes as part of storing data into addressable fast write storage and/or as part of writing data into allocation units associated with erase blocks. Erasure codes may be used across storage devices, as well as within erase blocks or allocation units, or within and across Flash memory devices on a single storage device, to provide redundancy against single or multiple storage device failures or to protect against internal corruptions of Flash memory pages resulting from Flash memory operations or from degradation of Flash memory cells. Mirroring and erasure coding at various levels may be used to recover from multiple types of failures that occur separately or in combination.
The embodiments depicted with reference to FIGS. 2A-G illustrate a storage cluster that stores user data, such as user data originating from one or more user or client systems or other sources external to the storage cluster. The storage cluster distributes user data across storage nodes housed within a chassis, or across multiple chassis, using erasure coding and redundant copies of metadata. Erasure coding refers to a method of data protection or reconstruction in which data is stored across a set of different locations, such as disks, storage nodes or geographic locations. Flash memory is one type of solid-state memory that may be integrated with the embodiments, although the embodiments may be extended to other types of solid-state memory or other storage medium, including non-solid state memory. Control of storage locations and workloads are distributed across the storage locations in a clustered peer-to-peer system. Tasks such as mediating communications between the various storage nodes, detecting when a storage node has become unavailable, and balancing I/Os (inputs and outputs) across the various storage nodes, are all handled on a distributed basis. Data is laid out or distributed across multiple storage nodes in data fragments or stripes that support data recovery in some embodiments. Ownership of data can be reassigned within a cluster, independent of input and output patterns. This architecture described in more detail below allows a storage node in the cluster to fail, with the system remaining operational, since the data can be reconstructed from other storage nodes and thus remain available for input and output operations. In various embodiments, a storage node may be referred to as a cluster node, a blade, or a server.
The storage cluster may be contained within a chassis, i.e., an enclosure housing one or more storage nodes. A mechanism to provide power to each storage node, such as a power distribution bus, and a communication mechanism, such as a communication bus that enables communication between the storage nodes are included within the chassis. The storage cluster can run as an independent system in one location according to some embodiments. In one embodiment, a chassis contains at least two instances of both the power distribution and the communication bus which may be enabled or disabled independently. The internal communication bus may be an Ethernet bus, however, other technologies such as PCIe, InfiniBand, and others, are equally suitable. The chassis provides a port for an external communication bus for enabling communication between multiple chassis, directly or through a switch, and with client systems. The external communication may use a technology such as Ethernet, InfiniBand, Fibre Channel, etc. In some embodiments, the external communication bus uses different communication bus technologies for inter-chassis and client communication. If a switch is deployed within or between chassis, the switch may act as a translation between multiple protocols or technologies. When multiple chassis are connected to define a storage cluster, the storage cluster may be accessed by a client using either proprietary interfaces or standard interfaces such as network file system (‘NFS’), common internet file system (‘CIFS’), small computer system interface (‘SCSI’) or hypertext transfer protocol (‘HTTP’). Translation from the client protocol may occur at the switch, chassis external communication bus or within each storage node. In some embodiments, multiple chassis may be coupled or connected to each other through an aggregator switch. A portion and/or all of the coupled or connected chassis may be designated as a storage cluster. As discussed above, each chassis can have multiple blades, each blade has a media access control (‘MAC’) address, but the storage cluster is presented to an external network as having a single cluster IP address and a single MAC address in some embodiments.
Each storage node may be one or more storage servers and each storage server is connected to one or more non-volatile solid state memory units, which may be referred to as storage units or storage devices. One embodiment includes a single storage server in each storage node and between one to eight non-volatile solid state memory units, however this one example is not meant to be limiting. The storage server may include a processor, DRAM and interfaces for the internal communication bus and power distribution for each of the power buses. Inside the storage node, the interfaces and storage unit share a communication bus, e.g., PCI Express, in some embodiments. The non-volatile solid state memory units may directly access the internal communication bus interface through a storage node communication bus, or request the storage node to access the bus interface. The non-volatile solid state memory unit contains an embedded CPU, solid state storage controller, and a quantity of solid state mass storage, e.g., between 2-32 terabytes (‘TB’) in some embodiments. An embedded volatile storage medium, such as DRAM, and an energy reserve apparatus are included in the non-volatile solid state memory unit. In some embodiments, the energy reserve apparatus is a capacitor, super-capacitor, or battery that enables transferring a subset of DRAM contents to a stable storage medium in the case of power loss. In some embodiments, the non-volatile solid state memory unit is constructed with a storage class memory, such as phase change or magnetoresistive random access memory (‘MRAM’) that substitutes for DRAM and enables a reduced power hold-up apparatus.
One of many features of the storage nodes and non-volatile solid state storage is the ability to proactively rebuild data in a storage cluster. The storage nodes and non-volatile solid state storage can determine when a storage node or non-volatile solid state storage in the storage cluster is unreachable, independent of whether there is an attempt to read data involving that storage node or non-volatile solid state storage. The storage nodes and non-volatile solid state storage then cooperate to recover and rebuild the data in at least partially new locations. This constitutes a proactive rebuild, in that the system rebuilds data without waiting until the data is needed for a read access initiated from a client system employing the storage cluster. These and further details of the storage memory and operation thereof are discussed below.
FIG. 2A is a perspective view of a storage cluster 161, with multiple storage nodes 150 and internal solid-state memory coupled to each storage node to provide network attached storage or storage area network, in accordance with some embodiments. A network attached storage, storage area network, or a storage cluster, or other storage memory, could include one or more storage clusters 161, each having one or more storage nodes 150, in a flexible and reconfigurable arrangement of both the physical components and the amount of storage memory provided thereby. The storage cluster 161 is designed to fit in a rack, and one or more racks can be set up and populated as desired for the storage memory. The storage cluster 161 has a chassis 138 having multiple slots 142. It should be appreciated that chassis 138 may be referred to as a housing, enclosure, or rack unit. In one embodiment, the chassis 138 has fourteen slots 142, although other numbers of slots are readily devised. For example, some embodiments have four slots, eight slots, sixteen slots, thirty-two slots, or other suitable number of slots. Each slot 142 can accommodate one storage node 150 in some embodiments. Chassis 138 includes flaps 148 that can be utilized to mount the chassis 138 on a rack. Fans 144 provide air circulation for cooling of the storage nodes 150 and components thereof, although other cooling components could be used, or an embodiment could be devised without cooling components. A switch fabric 146 couples storage nodes 150 within chassis 138 together and to a network for communication to the memory. In an embodiment depicted in herein, the slots 142 to the left of the switch fabric 146 and fans 144 are shown occupied by storage nodes 150, while the slots 142 to the right of the switch fabric 146 and fans 144 are empty and available for insertion of storage node 150 for illustrative purposes. This configuration is one example, and one or more storage nodes 150 could occupy the slots 142 in various further arrangements. The storage node arrangements need not be sequential or adjacent in some embodiments. Storage nodes 150 are hot pluggable, meaning that a storage node 150 can be inserted into a slot 142 in the chassis 138, or removed from a slot 142, without stopping or powering down the system. Upon insertion or removal of storage node 150 from slot 142, the system automatically reconfigures in order to recognize and adapt to the change. Reconfiguration, in some embodiments, includes restoring redundancy and/or rebalancing data or load.
Each storage node 150 can have multiple components. In the embodiment shown here, the storage node 150 includes a printed circuit board 159 populated by a CPU 156, i.e., processor, a memory 154 coupled to the CPU 156, and a non-volatile solid state storage 152 coupled to the CPU 156, although other mountings and/or components could be used in further embodiments. The memory 154 has instructions which are executed by the CPU 156 and/or data operated on by the CPU 156. As further explained below, the non-volatile solid state storage 152 includes flash or, in further embodiments, other types of solid-state memory.
Referring to FIG. 2A, storage cluster 161 is scalable, meaning that storage capacity with non-uniform storage sizes is readily added, as described above. One or more storage nodes 150 can be plugged into or removed from each chassis and the storage cluster self-configures in some embodiments. Plug-in storage nodes 150, whether installed in a chassis as delivered or later added, can have different sizes. For example, in one embodiment a storage node 150 can have any multiple of 4 TB, e.g., 8 TB, 12 TB, 16 TB, 32 TB, etc. In further embodiments, a storage node 150 could have any multiple of other storage amounts or capacities. Storage capacity of each storage node 150 is broadcast, and influences decisions of how to stripe the data. For maximum storage efficiency, an embodiment can self-configure as wide as possible in the stripe, subject to a predetermined requirement of continued operation with loss of up to one, or up to two, non-volatile solid state storage units 152 or storage nodes 150 within the chassis.
FIG. 2B is a block diagram showing a communications interconnect 171A-F and power distribution bus 172 coupling multiple storage nodes 150. Referring back to FIG. 2A, the communications interconnect 171A-F can be included in or implemented with the switch fabric 146 in some embodiments. Where multiple storage clusters 161 occupy a rack, the communications interconnect 171A-F can be included in or implemented with a top of rack switch, in some embodiments. As illustrated in FIG. 2B, storage cluster 161 is enclosed within a single chassis 138. External port 176 is coupled to storage nodes 150 through communications interconnect 171A-F, while external port 174 is coupled directly to a storage node. External power port 178 is coupled to power distribution bus 172. Storage nodes 150 may include varying amounts and differing capacities of non-volatile solid state storage 152 as described with reference to FIG. 2A. In addition, one or more storage nodes 150 may be a compute only storage node as illustrated in FIG. 2B. Authorities 168 are implemented on the non-volatile solid state storages 152, for example as lists or other data structures stored in memory. In some embodiments the authorities are stored within the non-volatile solid state storage 152 and supported by software executing on a controller or other processor of the non-volatile solid state storage 152. In a further embodiment, authorities 168 are implemented on the storage nodes 150, for example as lists or other data structures stored in the memory 154 and supported by software executing on the CPU 156 of the storage node 150. Authorities 168 control how and where data is stored in the non-volatile solid state storages 152 in some embodiments. This control assists in determining which type of erasure coding scheme is applied to the data, and which storage nodes 150 have which portions of the data. Each authority 168 may be assigned to a non-volatile solid state storage 152. Each authority may control a range of inode numbers, segment numbers, or other data identifiers which are assigned to data by a file system, by the storage nodes 150, or by the non-volatile solid state storage 152, in various embodiments.
Every piece of data, and every piece of metadata, has redundancy in the system in some embodiments. In addition, every piece of data and every piece of metadata has an owner, which may be referred to as an authority. If that authority is unreachable, for example through failure of a storage node, there is a plan of succession for how to find that data or that metadata. In various embodiments, there are redundant copies of authorities 168. Authorities 168 have a relationship to storage nodes 150 and non-volatile solid state storage 152 in some embodiments. Each authority 168, covering a range of data segment numbers or other identifiers of the data, may be assigned to a specific non-volatile solid state storage 152. In some embodiments the authorities 168 for all of such ranges are distributed over the non-volatile solid state storages 152 of a storage cluster. Each storage node 150 has a network port that provides access to the non-volatile solid state storage(s) 152 of that storage node 150. Data can be stored in a segment, which is associated with a segment number and that segment number is an indirection for a configuration of a RAID (redundant array of independent disks) stripe in some embodiments. The assignment and use of the authorities 168 thus establishes an indirection to data. Indirection may be referred to as the ability to reference data indirectly, in this case via an authority 168, in accordance with some embodiments. A segment identifies a set of non-volatile solid state storage 152 and a local identifier into the set of non-volatile solid state storage 152 that may contain data. In some embodiments, the local identifier is an offset into the device and may be reused sequentially by multiple segments. In other embodiments the local identifier is unique for a specific segment and never reused. The offsets in the non-volatile solid state storage 152 are applied to locating data for writing to or reading from the non-volatile solid state storage 152 (in the form of a RAID stripe). Data is striped across multiple units of non-volatile solid state storage 152, which may include or be different from the non-volatile solid state storage 152 having the authority 168 for a particular data segment.
If there is a change in where a particular segment of data is located, e.g., during a data move or a data reconstruction, the authority 168 for that data segment should be consulted, at that non-volatile solid state storage 152 or storage node 150 having that authority 168. In order to locate a particular piece of data, embodiments calculate a hash value for a data segment or apply an inode number or a data segment number. The output of this operation points to a non-volatile solid state storage 152 having the authority 168 for that particular piece of data. In some embodiments there are two stages to this operation. The first stage maps an entity identifier (ID), e.g., a segment number, inode number, or directory number to an authority identifier. This mapping may include a calculation such as a hash or a bit mask. The second stage is mapping the authority identifier to a particular non-volatile solid state storage 152, which may be done through an explicit mapping. The operation is repeatable, so that when the calculation is performed, the result of the calculation repeatably and reliably points to a particular non-volatile solid state storage 152 having that authority 168. The operation may include the set of reachable storage nodes as input. If the set of reachable non-volatile solid state storage units changes the optimal set changes. In some embodiments, the persisted value is the current assignment (which is always true) and the calculated value is the target assignment the cluster will attempt to reconfigure towards. This calculation may be used to determine the optimal non-volatile solid state storage 152 for an authority in the presence of a set of non-volatile solid state storage 152 that are reachable and constitute the same cluster. The calculation also determines an ordered set of peer non-volatile solid state storage 152 that will also record the authority to non-volatile solid state storage mapping so that the authority may be determined even if the assigned non-volatile solid state storage is unreachable. A duplicate or substitute authority 168 may be consulted if a specific authority 168 is unavailable in some embodiments.
With reference to FIGS. 2A and 2B, two of the many tasks of the CPU 156 on a storage node 150 are to break up write data, and reassemble read data. When the system has determined that data is to be written, the authority 168 for that data is located as above. When the segment ID for data is already determined the request to write is forwarded to the non-volatile solid state storage 152 currently determined to be the host of the authority 168 determined from the segment. The host CPU 156 of the storage node 150, on which the non-volatile solid state storage 152 and corresponding authority 168 reside, then breaks up or shards the data and transmits the data out to various non-volatile solid state storage 152. The transmitted data is written as a data stripe in accordance with an erasure coding scheme. In some embodiments, data is requested to be pulled, and in other embodiments, data is pushed. In reverse, when data is read, the authority 168 for the segment ID containing the data is located as described above. The host CPU 156 of the storage node 150 on which the non-volatile solid state storage 152 and corresponding authority 168 reside requests the data from the non-volatile solid state storage and corresponding storage nodes pointed to by the authority. In some embodiments the data is read from flash storage as a data stripe. The host CPU 156 of storage node 150 then reassembles the read data, correcting any errors (if present) according to the appropriate erasure coding scheme, and forwards the reassembled data to the network. In further embodiments, some or all of these tasks can be handled in the non-volatile solid state storage 152. In some embodiments, the segment host requests the data be sent to storage node 150 by requesting pages from storage and then sending the data to the storage node making the original request.
In some systems, for example in UNIX-style file systems, data is handled with an index node or inode, which specifies a data structure that represents an object in a file system. The object could be a file or a directory, for example. Metadata may accompany the object, as attributes such as permission data and a creation timestamp, among other attributes. A segment number could be assigned to all or a portion of such an object in a file system. In other systems, data segments are handled with a segment number assigned elsewhere. For purposes of discussion, the unit of distribution is an entity, and an entity can be a file, a directory or a segment. That is, entities are units of data or metadata stored by a storage system. Entities are grouped into sets called authorities. Each authority has an authority owner, which is a storage node that has the exclusive right to update the entities in the authority. In other words, a storage node contains the authority, and that the authority, in turn, contains entities.
A segment is a logical container of data in accordance with some embodiments. A segment is an address space between medium address space and physical flash locations, i.e., the data segment number, are in this address space. Segments may also contain meta-data, which enable data redundancy to be restored (rewritten to different flash locations or devices) without the involvement of higher level software. In one embodiment, an internal format of a segment contains client data and medium mappings to determine the position of that data. Each data segment is protected, e.g., from memory and other failures, by breaking the segment into a number of data and parity shards, where applicable. The data and parity shards are distributed, i.e., striped, across non-volatile solid state storage 152 coupled to the host CPUs 156 (See FIGS. 2E and 2G) in accordance with an erasure coding scheme. Usage of the term segments refers to the container and its place in the address space of segments in some embodiments. Usage of the term stripe refers to the same set of shards as a segment and includes how the shards are distributed along with redundancy or parity information in accordance with some embodiments.
A series of address-space transformations takes place across an entire storage system. At the top are the directory entries (file names) which link to an inode. Inodes point into medium address space, where data is logically stored. Medium addresses may be mapped through a series of indirect mediums to spread the load of large files, or implement data services like deduplication or snapshots. Medium addresses may be mapped through a series of indirect mediums to spread the load of large files, or implement data services like deduplication or snapshots. Segment addresses are then translated into physical flash locations. Physical flash locations have an address range bounded by the amount of flash in the system in accordance with some embodiments. Medium addresses and segment addresses are logical containers, and in some embodiments use a 128 bit or larger identifier so as to be practically infinite, with a likelihood of reuse calculated as longer than the expected life of the system. Addresses from logical containers are allocated in a hierarchical fashion in some embodiments. Initially, each non-volatile solid state storage unit 152 may be assigned a range of address space. Within this assigned range, the non-volatile solid state storage 152 is able to allocate addresses without synchronization with other non-volatile solid state storage 152.
Data and metadata is stored by a set of underlying storage layouts that are optimized for varying workload patterns and storage devices. These layouts incorporate multiple redundancy schemes, compression formats and index algorithms. Some of these layouts store information about authorities and authority masters, while others store file metadata and file data. The redundancy schemes include error correction codes that tolerate corrupted bits within a single storage device (such as a NAND flash chip), erasure codes that tolerate the failure of multiple storage nodes, and replication schemes that tolerate data center or regional failures. In some embodiments, low density parity check (‘LDPC’) code is used within a single storage unit. Reed-Solomon encoding is used within a storage cluster, and mirroring is used within a storage grid in some embodiments. Metadata may be stored using an ordered log structured index (such as a Log Structured Merge Tree), and large data may not be stored in a log structured layout.
In order to maintain consistency across multiple copies of an entity, the storage nodes agree implicitly on two things through calculations: (1) the authority that contains the entity, and (2) the storage node that contains the authority. The assignment of entities to authorities can be done by pseudo randomly assigning entities to authorities, by splitting entities into ranges based upon an externally produced key, or by placing a single entity into each authority. Examples of pseudorandom schemes are linear hashing and the Replication Under Scalable Hashing (‘RUSH’) family of hashes, including Controlled Replication Under Scalable Hashing (‘CRUSH’). In some embodiments, pseudo-random assignment is utilized only for assigning authorities to nodes because the set of nodes can change. The set of authorities cannot change so any subjective function may be applied in these embodiments. Some placement schemes automatically place authorities on storage nodes, while other placement schemes rely on an explicit mapping of authorities to storage nodes. In some embodiments, a pseudorandom scheme is utilized to map from each authority to a set of candidate authority owners. A pseudorandom data distribution function related to CRUSH may assign authorities to storage nodes and create a list of where the authorities are assigned. Each storage node has a copy of the pseudorandom data distribution function, and can arrive at the same calculation for distributing, and later finding or locating an authority. Each of the pseudorandom schemes requires the reachable set of storage nodes as input in some embodiments in order to conclude the same target nodes. Once an entity has been placed in an authority, the entity may be stored on physical devices so that no expected failure will lead to unexpected data loss. In some embodiments, rebalancing algorithms attempt to store the copies of all entities within an authority in the same layout and on the same set of machines.
Examples of expected failures include device failures, stolen machines, datacenter fires, and regional disasters, such as nuclear or geological events. Different failures lead to different levels of acceptable data loss. In some embodiments, a stolen storage node impacts neither the security nor the reliability of the system, while depending on system configuration, a regional event could lead to no loss of data, a few seconds or minutes of lost updates, or even complete data loss.
In the embodiments, the placement of data for storage redundancy is independent of the placement of authorities for data consistency. In some embodiments, storage nodes that contain authorities do not contain any persistent storage. Instead, the storage nodes are connected to non-volatile solid state storage units that do not contain authorities. The communications interconnect between storage nodes and non-volatile solid state storage units consists of multiple communication technologies and has non-uniform performance and fault tolerance characteristics. In some embodiments, as mentioned above, non-volatile solid state storage units are connected to storage nodes via PCI express, storage nodes are connected together within a single chassis using Ethernet backplane, and chassis are connected together to form a storage cluster. Storage clusters are connected to clients using Ethernet or fiber channel in some embodiments. If multiple storage clusters are configured into a storage grid, the multiple storage clusters are connected using the Internet or other long-distance networking links, such as a “metro scale” link or private link that does not traverse the internet.
Authority owners have the exclusive right to modify entities, to migrate entities from one non-volatile solid state storage unit to another non-volatile solid state storage unit, and to add and remove copies of entities. This allows for maintaining the redundancy of the underlying data. When an authority owner fails, is going to be decommissioned, or is overloaded, the authority is transferred to a new storage node. Transient failures make it non-trivial to ensure that all non-faulty machines agree upon the new authority location. The ambiguity that arises due to transient failures can be achieved automatically by a consensus protocol such as Paxos, hot-warm failover schemes, via manual intervention by a remote system administrator, or by a local hardware administrator (such as by physically removing the failed machine from the cluster, or pressing a button on the failed machine). In some embodiments, a consensus protocol is used, and failover is automatic. If too many failures or replication events occur in too short a time period, the system goes into a self-preservation mode and halts replication and data movement activities until an administrator intervenes in accordance with some embodiments.
As authorities are transferred between storage nodes and authority owners update entities in their authorities, the system transfers messages between the storage nodes and non-volatile solid state storage units. With regard to persistent messages, messages that have different purposes are of different types. Depending on the type of the message, the system maintains different ordering and durability guarantees. As the persistent messages are being processed, the messages are temporarily stored in multiple durable and non-durable storage hardware technologies. In some embodiments, messages are stored in RAM, NVRAM and on NAND flash devices, and a variety of protocols are used in order to make efficient use of each storage medium. Latency-sensitive client requests may be persisted in replicated NVRAM, and then later NAND, while background rebalancing operations are persisted directly to NAND.
Persistent messages are persistently stored prior to being transmitted. This allows the system to continue to serve client requests despite failures and component replacement. Although many hardware components contain unique identifiers that are visible to system administrators, manufacturer, hardware supply chain and ongoing monitoring quality control infrastructure, applications running on top of the infrastructure address virtualize addresses. These virtualized addresses do not change over the lifetime of the storage system, regardless of component failures and replacements. This allows each component of the storage system to be replaced over time without reconfiguration or disruptions of client request processing, i.e., the system supports non-disruptive upgrades.
In some embodiments, the virtualized addresses are stored with sufficient redundancy. A continuous monitoring system correlates hardware and software status and the hardware identifiers. This allows detection and prediction of failures due to faulty components and manufacturing details. The monitoring system also enables the proactive transfer of authorities and entities away from impacted devices before failure occurs by removing the component from the critical path in some embodiments.
FIG. 2C is a multiple level block diagram, showing contents of a storage node 150 and contents of a non-volatile solid state storage 152 of the storage node 150. Data is communicated to and from the storage node 150 by a network interface controller (‘NIC’) 202 in some embodiments. Each storage node 150 has a CPU 156, and one or more non-volatile solid state storage 152, as discussed above. Moving down one level in FIG. 2C, each non-volatile solid state storage 152 has a relatively fast non-volatile solid state memory, such as nonvolatile random access memory (‘NVRAM’) 204, and flash memory 206. In some embodiments, NVRAM 204 may be a component that does not require program/erase cycles (DRAM, MRAM, PCM), and can be a memory that can support being written vastly more often than the memory is read from. Moving down another level in FIG. 2C, the NVRAM 204 is implemented in one embodiment as high speed volatile memory, such as dynamic random access memory (DRAM) 216, backed up by energy reserve 218. Energy reserve 218 provides sufficient electrical power to keep the DRAM 216 powered long enough for contents to be transferred to the flash memory 206 in the event of power failure. In some embodiments, energy reserve 218 is a capacitor, super-capacitor, battery, or other device, that supplies a suitable supply of energy sufficient to enable the transfer of the contents of DRAM 216 to a stable storage medium in the case of power loss. The flash memory 206 is implemented as multiple flash dies 222, which may be referred to as packages of flash dies 222 or an array of flash dies 222. It should be appreciated that the flash dies 222 could be packaged in any number of ways, with a single die per package, multiple dies per package (i.e. multichip packages), in hybrid packages, as bare dies on a printed circuit board or other substrate, as encapsulated dies, etc. In the embodiment shown, the non-volatile solid state storage 152 has a controller 212 or other processor, and an input output (I/O) port 210 coupled to the controller 212. I/O port 210 is coupled to the CPU 156 and/or the network interface controller 202 of the flash storage node 150. Flash input output (I/O) port 220 is coupled to the flash dies 222, and a direct memory access unit (DMA) 214 is coupled to the controller 212, the DRAM 216 and the flash dies 222. In the embodiment shown, the I/O port 210, controller 212, DMA unit 214 and flash I/O port 220 are implemented on a programmable logic device (‘PLD’) 208, e.g., a field programmable gate array (FPGA). In this embodiment, each flash die 222 has pages, organized as sixteen kB (kilobyte) pages 224, and a register 226 through which data can be written to or read from the flash die 222. In further embodiments, other types of solid-state memory are used in place of, or in addition to flash memory illustrated within flash die 222.
Storage clusters 161, in various embodiments as disclosed herein, can be contrasted with storage arrays in general. The storage nodes 150 are part of a collection that creates the storage cluster 161. Each storage node 150 owns a slice of data and computing required to provide the data. Multiple storage nodes 150 cooperate to store and retrieve the data. Storage memory or storage devices, as used in storage arrays in general, are less involved with processing and manipulating the data. Storage memory or storage devices in a storage array receive commands to read, write, or erase data. The storage memory or storage devices in a storage array are not aware of a larger system in which they are embedded, or what the data means. Storage memory or storage devices in storage arrays can include various types of storage memory, such as RAM, solid state drives, hard disk drives, etc. The storage units 152 described herein have multiple interfaces active simultaneously and serving multiple purposes. In some embodiments, some of the functionality of a storage node 150 is shifted into a storage unit 152, transforming the storage unit 152 into a combination of storage unit 152 and storage node 150. Placing computing (relative to storage data) into the storage unit 152 places this computing closer to the data itself. The various system embodiments have a hierarchy of storage node layers with different capabilities. By contrast, in a storage array, a controller owns and knows everything about all of the data that the controller manages in a shelf or storage devices. In a storage cluster 161, as described herein, multiple controllers in multiple storage units 152 and/or storage nodes 150 cooperate in various ways (e.g., for erasure coding, data sharding, metadata communication and redundancy, storage capacity expansion or contraction, data recovery, and so on).
FIG. 2D shows a storage server environment, which uses embodiments of the storage nodes 150 and storage units 152 of FIGS. 2A-C. In this version, each storage unit 152 has a processor such as controller 212 (see FIG. 2C), an FPGA (field programmable gate array), flash memory 206, and NVRAM 204 (which is super-capacitor backed DRAM 216, see FIGS. 2B and 2C) on a PCIe (peripheral component interconnect express) board in a chassis 138 (see FIG. 2A). The storage unit 152 may be implemented as a single board containing storage, and may be the largest tolerable failure domain inside the chassis. In some embodiments, up to two storage units 152 may fail and the device will continue with no data loss.
The physical storage is divided into named regions based on application usage in some embodiments. The NVRAM 204 is a contiguous block of reserved memory in the storage unit 152 DRAM 216, and is backed by NAND flash. NVRAM 204 is logically divided into multiple memory regions written for two as spool (e.g., spool_region). Space within the NVRAM 204 spools is managed by each authority 168 independently. Each device provides an amount of storage space to each authority 168. That authority 168 further manages lifetimes and allocations within that space. Examples of a spool include distributed transactions or notions. When the primary power to a storage unit 152 fails, onboard super-capacitors provide a short duration of power hold up. During this holdup interval, the contents of the NVRAM 204 are flushed to flash memory 206. On the next power-on, the contents of the NVRAM 204 are recovered from the flash memory 206.
As for the storage unit controller, the responsibility of the logical “controller” is distributed across each of the blades containing authorities 168. This distribution of logical control is shown in FIG. 2D as a host controller 242, mid-tier controller 244 and storage unit controller(s) 246. Management of the control plane and the storage plane are treated independently, although parts may be physically co-located on the same blade. Each authority 168 effectively serves as an independent controller. Each authority 168 provides its own data and metadata structures, its own background workers, and maintains its own lifecycle.
FIG. 2E is a blade 252 hardware block diagram, showing a control plane 254, compute and storage planes 256, 258, and authorities 168 interacting with underlying physical resources, using embodiments of the storage nodes 150 and storage units 152 of FIGS. 2A-C in the storage server environment of FIG. 2D. The control plane 254 is partitioned into a number of authorities 168 which can use the compute resources in the compute plane 256 to run on any of the blades 252. The storage plane 258 is partitioned into a set of devices, each of which provides access to flash 206 and NVRAM 204 resources.
In the compute and storage planes 256, 258 of FIG. 2E, the authorities 168 interact with the underlying physical resources (i.e., devices). From the point of view of an authority 168, its resources are striped over all of the physical devices. From the point of view of a device, it provides resources to all authorities 168, irrespective of where the authorities happen to run. Each authority 168 has allocated or has been allocated one or more partitions 260 of storage memory in the storage units 152, e.g. partitions 260 in flash memory 206 and NVRAM 204. Each authority 168 uses those allocated partitions 260 that belong to it, for writing or reading user data. Authorities can be associated with differing amounts of physical storage of the system. For example, one authority 168 could have a larger number of partitions 260 or larger sized partitions 260 in one or more storage units 152 than one or more other authorities 168.
FIG. 2F depicts elasticity software layers in blades 252 of a storage cluster, in accordance with some embodiments. In the elasticity structure, elasticity software is symmetric, i.e., each blade's compute module 270 runs the three identical layers of processes depicted in FIG. 2F. Storage managers 274 execute read and write requests from other blades 252 for data and metadata stored in local storage unit 152 NVRAM 204 and flash 206. Authorities 168 fulfill client requests by issuing the necessary reads and writes to the blades 252 on whose storage units 152 the corresponding data or metadata resides. Endpoints 272 parse client connection requests received from switch fabric 146 supervisory software, relay the client connection requests to the authorities 168 responsible for fulfillment, and relay the authorities' 168 responses to clients. The symmetric three-layer structure enables the storage system's high degree of concurrency. Elasticity scales out efficiently and reliably in these embodiments. In addition, elasticity implements a unique scale-out technique that balances work evenly across all resources regardless of client access pattern, and maximizes concurrency by eliminating much of the need for inter-blade coordination that typically occurs with conventional distributed locking.
Still referring to FIG. 2F, authorities 168 running in the compute modules 270 of a blade 252 perform the internal operations required to fulfill client requests. One feature of elasticity is that authorities 168 are stateless, i.e., they cache active data and metadata in their own blades' 252 DRAMs for fast access, but the authorities store every update in their NVRAM 204 partitions on three separate blades 252 until the update has been written to flash 206. All the storage system writes to NVRAM 204 are in triplicate to partitions on three separate blades 252 in some embodiments. With triple-mirrored NVRAM 204 and persistent storage protected by parity and Reed-Solomon RAID checksums, the storage system can survive concurrent failure of two blades 252 with no loss of data, metadata, or access to either.
Because authorities 168 are stateless, they can migrate between blades 252. Each authority 168 has a unique identifier. NVRAM 204 and flash 206 partitions are associated with authorities' 168 identifiers, not with the blades 252 on which they are running in some. Thus, when an authority 168 migrates, the authority 168 continues to manage the same storage partitions from its new location. When a new blade 252 is installed in an embodiment of the storage cluster, the system automatically rebalances load by: partitioning the new blade's 252 storage for use by the system's authorities 168, migrating selected authorities 168 to the new blade 252, starting endpoints 272 on the new blade 252 and including them in the switch fabric's 146 client connection distribution algorithm.
From their new locations, migrated authorities 168 persist the contents of their NVRAM 204 partitions on flash 206, process read and write requests from other authorities 168, and fulfill the client requests that endpoints 272 direct to them. Similarly, if a blade 252 fails or is removed, the system redistributes its authorities 168 among the system's remaining blades 252. The redistributed authorities 168 continue to perform their original functions from their new locations.
FIG. 2G depicts authorities 168 and storage resources in blades 252 of a storage cluster, in accordance with some embodiments. Each authority 168 is exclusively responsible for a partition of the flash 206 and NVRAM 204 on each blade 252. The authority 168 manages the content and integrity of its partitions independently of other authorities 168. Authorities 168 compress incoming data and preserve it temporarily in their NVRAM 204 partitions, and then consolidate, RAID-protect, and persist the data in segments of the storage in their flash 206 partitions. As the authorities 168 write data to flash 206, storage managers 274 perform the necessary flash translation to optimize write performance and maximize media longevity. In the background, authorities 168 “garbage collect,” or reclaim space occupied by data that clients have made obsolete by overwriting the data. It should be appreciated that since authorities' 168 partitions are disjoint, there is no need for distributed locking to execute client and writes or to perform background functions.
The embodiments described herein may utilize various software, communication and/or networking protocols. In addition, the configuration of the hardware and/or software may be adjusted to accommodate various protocols. For example, the embodiments may utilize Active Directory, which is a database based system that provides authentication, directory, policy, and other services in a WINDOWS™ environment. In these embodiments, LDAP (Lightweight Directory Access Protocol) is one example application protocol for querying and modifying items in directory service providers such as Active Directory. In some embodiments, a network lock manager (‘NLM’) is utilized as a facility that works in cooperation with the Network File System (‘NFS’) to provide a System V style of advisory file and record locking over a network. The Server Message Block (‘SMB’) protocol, one version of which is also known as Common Internet File System (‘CIFS’), may be integrated with the storage systems discussed herein. SMP operates as an application-layer network protocol typically used for providing shared access to files, printers, and serial ports and miscellaneous communications between nodes on a network. SMB also provides an authenticated inter-process communication mechanism. AMAZON™ S3 (Simple Storage Service) is a web service offered by Amazon Web Services, and the systems described herein may interface with Amazon S3 through web services interfaces (REST (representational state transfer), SOAP (simple object access protocol), and BitTorrent). A RESTful API (application programming interface) breaks down a transaction to create a series of small modules. Each module addresses a particular underlying part of the transaction. The control or permissions provided with these embodiments, especially for object data, may include utilization of an access control list (‘ACL’). The ACL is a list of permissions attached to an object and the ACL specifies which users or system processes are granted access to objects, as well as what operations are allowed on given objects. The systems may utilize Internet Protocol version 6 (‘IPv6’), as well as IPv4, for the communications protocol that provides an identification and location system for computers on networks and routes traffic across the Internet. The routing of packets between networked systems may include Equal-cost multi-path routing (‘ECMP’), which is a routing strategy where next-hop packet forwarding to a single destination can occur over multiple “best paths” which tie for top place in routing metric calculations. Multi-path routing can be used in conjunction with most routing protocols, because it is a per-hop decision limited to a single router. The software may support Multi-tenancy, which is an architecture in which a single instance of a software application serves multiple customers. Each customer may be referred to as a tenant. Tenants may be given the ability to customize some parts of the application, but may not customize the application's code, in some embodiments. The embodiments may maintain audit logs. An audit log is a document that records an event in a computing system. In addition to documenting what resources were accessed, audit log entries typically include destination and source addresses, a timestamp, and user login information for compliance with various regulations. The embodiments may support various key management policies, such as encryption key rotation. In addition, the system may support dynamic root passwords or some variation dynamically changing passwords.
FIG. 3A sets forth a diagram of a storage system 306 that is coupled for data communications with a cloud services provider 302 in accordance with some embodiments of the present disclosure. Although depicted in less detail, the storage system 306 depicted in FIG. 3A may be similar to the storage systems described above with reference to FIGS. 1A-1D and FIGS. 2A-2G. In some embodiments, the storage system 306 depicted in FIG. 3A may be embodied as a storage system that includes imbalanced active/active controllers, as a storage system that includes balanced active/active controllers, as a storage system that includes active/active controllers where less than all of each controller's resources are utilized such that each controller has reserve resources that may be used to support failover, as a storage system that includes fully active/active controllers, as a storage system that includes dataset-segregated controllers, as a storage system that includes dual-layer architectures with front-end controllers and back-end integrated storage controllers, as a storage system that includes scale-out clusters of dual-controller arrays, as well as combinations of such embodiments.
In the example depicted in FIG. 3A, the storage system 306 is coupled to the cloud services provider 302 via a data communications link 304. The data communications link 304 may be embodied as a dedicated data communications link, as a data communications pathway that is provided through the use of one or data communications networks such as a wide area network (‘WAN’) or local area network (‘LAN’), or as some other mechanism capable of transporting digital information between the storage system 306 and the cloud services provider 302. Such a data communications link 304 may be fully wired, fully wireless, or some aggregation of wired and wireless data communications pathways. In such an example, digital information may be exchanged between the storage system 306 and the cloud services provider 302 via the data communications link 304 using one or more data communications protocols. For example, digital information may be exchanged between the storage system 306 and the cloud services provider 302 via the data communications link 304 using the handheld device transfer protocol (‘HDTP’), hypertext transfer protocol (‘HTTP’), internet protocol (‘IP’), real-time transfer protocol (‘RTP’), transmission control protocol (‘TCP’), user datagram protocol (‘UDP’), wireless application protocol (‘WAP’), or other protocol.
The cloud services provider 302 depicted in FIG. 3A may be embodied, for example, as a system and computing environment that provides services to users of the cloud services provider 302 through the sharing of computing resources via the data communications link 304. The cloud services provider 302 may provide on-demand access to a shared pool of configurable computing resources such as computer networks, servers, storage, applications and services, and so on. The shared pool of configurable resources may be rapidly provisioned and released to a user of the cloud services provider 302 with minimal management effort. Generally, the user of the cloud services provider 302 is unaware of the exact computing resources utilized by the cloud services provider 302 to provide the services. Although in many cases such a cloud services provider 302 may be accessible via the Internet, readers of skill in the art will recognize that any system that abstracts the use of shared resources to provide services to a user through any data communications link may be considered a cloud services provider 302.
In the example depicted in FIG. 3A, the cloud services provider 302 may be configured to provide a variety of services to the storage system 306 and users of the storage system 306 through the implementation of various service models. For example, the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of an infrastructure as a service (‘IaaS’) service model where the cloud services provider 302 offers computing infrastructure such as virtual machines and other resources as a service to subscribers. In addition, the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of a platform as a service (‘PaaS’) service model where the cloud services provider 302 offers a development environment to application developers. Such a development environment may include, for example, an operating system, programming-language execution environment, database, web server, or other components that may be utilized by application developers to develop and run software solutions on a cloud platform. Furthermore, the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of a software as a service (‘SaaS’) service model where the cloud services provider 302 offers application software, databases, as well as the platforms that are used to run the applications to the storage system 306 and users of the storage system 306, providing the storage system 306 and users of the storage system 306 with on-demand software and eliminating the need to install and run the application on local computers, which may simplify maintenance and support of the application. The cloud services provider 302 may be further configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of an authentication as a service (‘AaaS’) service model where the cloud services provider 302 offers authentication services that can be used to secure access to applications, data sources, or other resources. The cloud services provider 302 may also be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of a storage as a service model where the cloud services provider 302 offers access to its storage infrastructure for use by the storage system 306 and users of the storage system 306. Readers will appreciate that the cloud services provider 302 may be configured to provide additional services to the storage system 306 and users of the storage system 306 through the implementation of additional service models, as the service models described above are included only for explanatory purposes and in no way represent a limitation of the services that may be offered by the cloud services provider 302 or a limitation as to the service models that may be implemented by the cloud services provider 302.
In the example depicted in FIG. 3A, the cloud services provider 302 may be embodied, for example, as a private cloud, as a public cloud, or as a combination of a private cloud and public cloud. In an embodiment in which the cloud services provider 302 is embodied as a private cloud, the cloud services provider 302 may be dedicated to providing services to a single organization rather than providing services to multiple organizations. In an embodiment where the cloud services provider 302 is embodied as a public cloud, the cloud services provider 302 may provide services to multiple organizations. Public cloud and private cloud deployment models may differ and may come with various advantages and disadvantages. For example, because a public cloud deployment involves the sharing of a computing infrastructure across different organization, such a deployment may not be ideal for organizations with security concerns, mission-critical workloads, uptime requirements demands, and so on. While a private cloud deployment can address some of these issues, a private cloud deployment may require on-premises staff to manage the private cloud. In still alternative embodiments, the cloud services provider 302 may be embodied as a mix of a private and public cloud services with a hybrid cloud deployment.
Although not explicitly depicted in FIG. 3A, readers will appreciate that additional hardware components and additional software components may be necessary to facilitate the delivery of cloud services to the storage system 306 and users of the storage system 306. For example, the storage system 306 may be coupled to (or even include) a cloud storage gateway. Such a cloud storage gateway may be embodied, for example, as hardware-based or software-based appliance that is located on premise with the storage system 306. Such a cloud storage gateway may operate as a bridge between local applications that are executing on the storage array 306 and remote, cloud-based storage that is utilized by the storage array 306. Through the use of a cloud storage gateway, organizations may move primary iSCSI or NAS to the cloud services provider 302, thereby enabling the organization to save space on their on-premises storage systems. Such a cloud storage gateway may be configured to emulate a disk array, a block-based device, a file server, or other storage system that can translate the SCSI commands, file server commands, or other appropriate command into REST-space protocols that facilitate communications with the cloud services provider 302.
In order to enable the storage system 306 and users of the storage system 306 to make use of the services provided by the cloud services provider 302, a cloud migration process may take place during which data, applications, or other elements from an organization's local systems (or even from another cloud environment) are moved to the cloud services provider 302. In order to successfully migrate data, applications, or other elements to the cloud services provider's 302 environment, middleware such as a cloud migration tool may be utilized to bridge gaps between the cloud services provider's 302 environment and an organization's environment. Such cloud migration tools may also be configured to address potentially high network costs and long transfer times associated with migrating large volumes of data to the cloud services provider 302, as well as addressing security concerns associated with sensitive data to the cloud services provider 302 over data communications networks. In order to further enable the storage system 306 and users of the storage system 306 to make use of the services provided by the cloud services provider 302, a cloud orchestrator may also be used to arrange and coordinate automated tasks in pursuit of creating a consolidated process or workflow. Such a cloud orchestrator may perform tasks such as configuring various components, whether those components are cloud components or on-premises components, as well as managing the interconnections between such components. The cloud orchestrator can simplify the inter-component communication and connections to ensure that links are correctly configured and maintained.
In the example depicted in FIG. 3A, and as described briefly above, the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the usage of a SaaS service model where the cloud services provider 302 offers application software, databases, as well as the platforms that are used to run the applications to the storage system 306 and users of the storage system 306, providing the storage system 306 and users of the storage system 306 with on-demand software and eliminating the need to install and run the application on local computers, which may simplify maintenance and support of the application. Such applications may take many forms in accordance with various embodiments of the present disclosure. For example, the cloud services provider 302 may be configured to provide access to data analytics applications to the storage system 306 and users of the storage system 306. Such data analytics applications may be configured, for example, to receive telemetry data phoned home by the storage system 306. Such telemetry data may describe various operating characteristics of the storage system 306 and may be analyzed, for example, to determine the health of the storage system 306, to identify workloads that are executing on the storage system 306, to predict when the storage system 306 will run out of various resources, to recommend configuration changes, hardware or software upgrades, workflow migrations, or other actions that may improve the operation of the storage system 306.
The cloud services provider 302 may also be configured to provide access to virtualized computing environments to the storage system 306 and users of the storage system 306. Such virtualized computing environments may be embodied, for example, as a virtual machine or other virtualized computer hardware platforms, virtual storage devices, virtualized computer network resources, and so on. Examples of such virtualized environments can include virtual machines that are created to emulate an actual computer, virtualized desktop environments that separate a logical desktop from a physical machine, virtualized file systems that allow uniform access to different types of concrete file systems, and many others.
For further explanation, FIG. 3B sets forth a diagram of a storage system 306 in accordance with some embodiments of the present disclosure. Although depicted in less detail, the storage system 306 depicted in FIG. 3B may be similar to the storage systems described above with reference to FIGS. 1A-1D and FIGS. 2A-2G as the storage system may include many of the components described above.
The storage system 306 depicted in FIG. 3B may include storage resources 308, which may be embodied in many forms. For example, in some embodiments the storage resources 308 can include nano-RAM or another form of nonvolatile random access memory that utilizes carbon nanotubes deposited on a substrate. In some embodiments, the storage resources 308 may include 3D crosspoint non-volatile memory in which bit storage is based on a change of bulk resistance, in conjunction with a stackable cross-gridded data access array. In some embodiments, the storage resources 308 may include flash memory, including single-level cell (‘SLC’) NAND flash, multi-level cell (‘MLC’) NAND flash, triple-level cell (‘TLC’) NAND flash, quad-level cell (‘QLC’) NAND flash, and others. In some embodiments, the storage resources 308 may include non-volatile magnetoresistive random-access memory (‘MRAM’), including spin transfer torque (‘STT’) MRAM, in which data is stored through the use of magnetic storage elements. In some embodiments, the example storage resources 308 may include non-volatile phase-change memory (‘PCM’) that may have the ability to hold multiple bits in a single cell as cells can achieve a number of distinct intermediary states. In some embodiments, the storage resources 308 may include quantum memory that allows for the storage and retrieval of photonic quantum information. In some embodiments, the example storage resources 308 may include resistive random-access memory (‘ReRAM’) in which data is stored by changing the resistance across a dielectric solid-state material. In some embodiments, the storage resources 308 may include storage class memory (‘SCM’) in which solid-state nonvolatile memory may be manufactured at a high density using some combination of sub-lithographic patterning techniques, multiple bits per cell, multiple layers of devices, and so on. Readers will appreciate that other forms of computer memories and storage devices may be utilized by the storage systems described above, including DRAM, SRAM, EEPROM, universal memory, and many others. The storage resources 308 depicted in FIG. 3A may be embodied in a variety of form factors, including but not limited to, dual in-line memory modules (‘DIMMs’), non-volatile dual in-line memory modules (‘NVDIMMs’), M.2, U.2, and others.
The example storage system 306 depicted in FIG. 3B may implement a variety of storage architectures. For example, storage systems in accordance with some embodiments of the present disclosure may utilize block storage where data is stored in blocks, and each block essentially acts as an individual hard drive. Storage systems in accordance with some embodiments of the present disclosure may utilize object storage, where data is managed as objects. Each object may include the data itself, a variable amount of metadata, and a globally unique identifier, where object storage can be implemented at multiple levels (e.g., device level, system level, interface level). Storage systems in accordance with some embodiments of the present disclosure utilize file storage in which data is stored in a hierarchical structure. Such data may be saved in files and folders, and presented to both the system storing it and the system retrieving it in the same format.
The example storage system 306 depicted in FIG. 3B may be embodied as a storage system in which additional storage resources can be added through the use of a scale-up model, additional storage resources can be added through the use of a scale-out model, or through some combination thereof. In a scale-up model, additional storage may be added by adding additional storage devices. In a scale-out model, however, additional storage nodes may be added to a cluster of storage nodes, where such storage nodes can include additional processing resources, additional networking resources, and so on.
The storage system 306 depicted in FIG. 3B also includes communications resources 310 that may be useful in facilitating data communications between components within the storage system 306, as well as data communications between the storage system 306 and computing devices that are outside of the storage system 306. The communications resources 310 may be configured to utilize a variety of different protocols and data communication fabrics to facilitate data communications between components within the storage systems as well as computing devices that are outside of the storage system. For example, the communications resources 310 can include fibre channel (‘FC’) technologies such as FC fabrics and FC protocols that can transport SCSI commands over FC networks. The communications resources 310 can also include FC over ethernet (‘FCoE’) technologies through which FC frames are encapsulated and transmitted over Ethernet networks. The communications resources 310 can also include InfiniBand (‘IB’) technologies in which a switched fabric topology is utilized to facilitate transmissions between channel adapters. The communications resources 310 can also include NVM Express (‘NVMe’) technologies and NVMe over fabrics (‘NVMeoF’) technologies through which non-volatile storage media attached via a PCI express (‘PCIe’) bus may be accessed. The communications resources 310 can also include mechanisms for accessing storage resources 308 within the storage system 306 utilizing serial attached SCSI (‘SAS’), serial ATA (‘SATA’) bus interfaces for connecting storage resources 308 within the storage system 306 to host bus adapters within the storage system 306, internet small computer systems interface (‘iSCSI’) technologies to provide block-level access to storage resources 308 within the storage system 306, and other communications resources that that may be useful in facilitating data communications between components within the storage system 306, as well as data communications between the storage system 306 and computing devices that are outside of the storage system 306.
The storage system 306 depicted in FIG. 3B also includes processing resources 312 that may be useful in useful in executing computer program instructions and performing other computational tasks within the storage system 306. The processing resources 312 may include one or more application-specific integrated circuits (‘ASICs’) that are customized for some particular purpose as well as one or more central processing units (‘CPUs’). The processing resources 312 may also include one or more digital signal processors (‘DSPs’), one or more field-programmable gate arrays (‘FPGAs’), one or more systems on a chip (‘SoCs’), or other form of processing resources 312. The storage system 306 may utilize the storage resources 312 to perform a variety of tasks including, but not limited to, supporting the execution of software resources 314 that will be described in greater detail below.
The storage system 306 depicted in FIG. 3B also includes software resources 314 that, when executed by processing resources 312 within the storage system 306, may perform various tasks. The software resources 314 may include, for example, one or more modules of computer program instructions that when executed by processing resources 312 within the storage system 306 are useful in carrying out various data protection techniques to preserve the integrity of data that is stored within the storage systems. Readers will appreciate that such data protection techniques may be carried out, for example, by system software executing on computer hardware within the storage system, by a cloud services provider, or in other ways. Such data protection techniques can include, for example, data archiving techniques that cause data that is no longer actively used to be moved to a separate storage device or separate storage system for long-term retention, data backup techniques through which data stored in the storage system may be copied and stored in a distinct location to avoid data loss in the event of equipment failure or some other form of catastrophe with the storage system, data replication techniques through which data stored in the storage system is replicated to another storage system such that the data may be accessible via multiple storage systems, data snapshotting techniques through which the state of data within the storage system is captured at various points in time, data and database cloning techniques through which duplicate copies of data and databases may be created, and other data protection techniques. Through the use of such data protection techniques, business continuity and disaster recovery objectives may be met as a failure of the storage system may not result in the loss of data stored in the storage system.
The software resources 314 may also include software that is useful in implementing software-defined storage (‘SDS’). In such an example, the software resources 314 may include one or more modules of computer program instructions that, when executed, are useful in policy-based provisioning and management of data storage that is independent of the underlying hardware. Such software resources 314 may be useful in implementing storage virtualization to separate the storage hardware from the software that manages the storage hardware.
The software resources 314 may also include software that is useful in facilitating and optimizing I/O operations that are directed to the storage resources 308 in the storage system 306. For example, the software resources 314 may include software modules that perform carry out various data reduction techniques such as, for example, data compression, data deduplication, and others. The software resources 314 may include software modules that intelligently group together I/O operations to facilitate better usage of the underlying storage resource 308, software modules that perform data migration operations to migrate from within a storage system, as well as software modules that perform other functions. Such software resources 314 may be embodied as one or more software containers or in many other ways.
Readers will appreciate that the various components depicted in FIG. 3B may be grouped into one or more optimized computing packages as converged infrastructures. Such converged infrastructures may include pools of computers, storage and networking resources that can be shared by multiple applications and managed in a collective manner using policy-driven processes. Such converged infrastructures may minimize compatibility issues between various components within the storage system 306 while also reducing various costs associated with the establishment and operation of the storage system 306. Such converged infrastructures may be implemented with a converged infrastructure reference architecture, with standalone appliances, with a software driven hyper-converged approach (e.g., hyper-converged infrastructures), or in other ways.
Readers will appreciate that the storage system 306 depicted in FIG. 3B may be useful for supporting various types of software applications. For example, the storage system 306 may be useful in supporting artificial intelligence (‘AI’) applications, database applications, DevOps projects, electronic design automation tools, event-driven software applications, high performance computing applications, simulation applications, high-speed data capture and analysis applications, machine learning applications, media production applications, media serving applications, picture archiving and communication systems (‘PACS’) applications, software development applications, virtual reality applications, augmented reality applications, and many other types of applications by providing storage resources to such applications.
The storage systems described above may operate to support a wide variety of applications. In view of the fact that the storage systems include compute resources, storage resources, and a wide variety of other resources, the storage systems may be well suited to support applications that are resource intensive such as, for example, AI applications. Such AI applications may enable devices to perceive their environment and take actions that maximize their chance of success at some goal. Examples of such AI applications can include IBM
Watson, Microsoft Oxford, Google DeepMind, Baidu Minwa, and others. The storage systems described above may also be well suited to support other types of applications that are resource intensive such as, for example, machine learning applications. Machine learning applications may perform various types of data analysis to automate analytical model building. Using algorithms that iteratively learn from data, machine learning applications can enable computers to learn without being explicitly programmed.
In addition to the resources already described, the storage systems described above may also include graphics processing units (‘GPUs’), occasionally referred to as visual processing unit (‘VPUs’). Such GPUs may be embodied as specialized electronic circuits that rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. Such GPUs may be included within any of the computing devices that are part of the storage systems described above, including as one of many individually scalable components of a storage system, where other examples of individually scalable components of such storage system can include storage components, memory components, compute components (e.g., CPUs, FPGAs, ASICs), networking components, software components, and others. In addition to GPUs, the storage systems described above may also include neural network processors (‘NNPs’) for use in various aspects of neural network processing. Such NNPs may be used in place of (or in addition to) GPUs and may be also be independently scalable.
As described above, the storage systems described herein may be configured to support artificial intelligence applications, machine learning applications, big data analytics applications, and many other types of applications. The rapid growth in these sorts of applications is being driven by three technologies: deep learning (DL), GPU processors, and Big Data. Deep learning is a computing model that makes use of massively parallel neural networks inspired by the human brain. Instead of experts handcrafting software, a deep learning model writes its own software by learning from lots of examples. A GPU is a modern processor with thousands of cores, well-suited to run algorithms that loosely represent the parallel nature of the human brain.
Advances in deep neural networks have ignited a new wave of algorithms and tools for data scientists to tap into their data with artificial intelligence (AI). With improved algorithms, larger data sets, and various frameworks (including open-source software libraries for machine learning across a range of tasks), data scientists are tackling new use cases like autonomous driving vehicles, natural language processing, and many others. Training deep neural networks, however, requires both high quality input data and large amounts of computation. GPUs are massively parallel processors capable of operating on large amounts of data simultaneously. When combined into a multi-GPU cluster, a high throughput pipeline may be required to feed input data from storage to the compute engines. Deep learning is more than just constructing and training models. There also exists an entire data pipeline that must be designed for the scale, iteration, and experimentation necessary for a data science team to succeed.
Data is the heart of modern AI and deep learning algorithms. Before training can begin, one problem that must be addressed revolves around collecting the labeled data that is crucial for training an accurate AI model. A full scale AI deployment may be required to continuously collect, clean, transform, label, and store large amounts of data. Adding additional high quality data points directly translates to more accurate models and better insights. Data samples may undergo a series of processing steps including, but not limited to: 1) ingesting the data from an external source into the training system and storing the data in raw form, 2) cleaning and transforming the data in a format convenient for training, including linking data samples to the appropriate label, 3) exploring parameters and models, quickly testing with a smaller dataset, and iterating to converge on the most promising models to push into the production cluster, 4) executing training phases to select random batches of input data, including both new and older samples, and feeding those into production GPU servers for computation to update model parameters, and 5) evaluating including using a holdback portion of the data not used in training in order to evaluate model accuracy on the holdout data. This lifecycle may apply for any type of parallelized machine learning, not just neural networks or deep learning. For example, standard machine learning frameworks may rely on CPUs instead of GPUs but the data ingest and training workflows may be the same. Readers will appreciate that a single shared storage data hub creates a coordination point throughout the lifecycle without the need for extra data copies among the ingest, preprocessing, and training stages. Rarely is the ingested data used for only one purpose, and shared storage gives the flexibility to train multiple different models or apply traditional analytics to the data.
Readers will appreciate that each stage in the AI data pipeline may have varying requirements from the data hub (e.g., the storage system or collection of storage systems). Scale-out storage systems must deliver uncompromising performance for all manner of access types and patterns—from small, metadata-heavy to large files, from random to sequential access patterns, and from low to high concurrency. The storage systems described above may serve as an ideal AI data hub as the systems may service unstructured workloads. In the first stage, data is ideally ingested and stored on to the same data hub that following stages will use, in order to avoid excess data copying. The next two steps can be done on a standard compute server that optionally includes a GPU, and then in the fourth and last stage, full training production jobs are run on powerful GPU-accelerated servers. Often, there is a production pipeline alongside an experimental pipeline operating on the same dataset. Further, the GPU-accelerated servers can be used independently for different models or joined together to train on one larger model, even spanning multiple systems for distributed training. If the shared storage tier is slow, then data must be copied to local storage for each phase, resulting in wasted time staging data onto different servers. The ideal data hub for the AI training pipeline delivers performance similar to data stored locally on the server node while also having the simplicity and performance to enable all pipeline stages to operate concurrently.
A data scientist works to improve the usefulness of the trained model through a wide variety of approaches: more data, better data, smarter training, and deeper models. In many cases, there will be teams of data scientists sharing the same datasets and working in parallel to produce new and improved training models. Often, there is a team of data scientists working within these phases concurrently on the same shared datasets. Multiple, concurrent workloads of data processing, experimentation, and full-scale training layer the demands of multiple access patterns on the storage tier. In other words, storage cannot just satisfy large file reads, but must contend with a mix of large and small file reads and writes. Finally, with multiple data scientists exploring datasets and models, it may be critical to store data in its native format to provide flexibility for each user to transform, clean, and use the data in a unique way. The storage systems described above may provide a natural shared storage home for the dataset, with data protection redundancy (e.g., by using RAID6) and the performance necessary to be a common access point for multiple developers and multiple experiments. Using the storage systems described above may avoid the need to carefully copy subsets of the data for local work, saving both engineering and GPU-accelerated servers use time. These copies become a constant and growing tax as the raw data set and desired transformations constantly update and change.
Readers will appreciate that a fundamental reason why deep learning has seen a surge in success is the continued improvement of models with larger data set sizes. In contrast, classical machine learning algorithms, like logistic regression, stop improving in accuracy at smaller data set sizes. As such, the separation of compute resources and storage resources may also allow independent scaling of each tier, avoiding many of the complexities inherent in managing both together. As the data set size grows or new data sets are considered, a scale out storage system must be able to expand easily. Similarly, if more concurrent training is required, additional GPUs or other compute resources can be added without concern for their internal storage. Furthermore, the storage systems described above may make building, operating, and growing an AI system easier due to the random read bandwidth provided by the storage systems, the ability to of the storage systems to randomly read small files (50 KB) high rates (meaning that no extra effort is required to aggregate individual data points to make larger, storage-friendly files), the ability of the storage systems to scale capacity and performance as either the dataset grows or the throughput requirements grow, the ability of the storage systems to support files or objects, the ability of the storage systems to tune performance for large or small files (i.e., no need for the user to provision filesystems), the ability of the storage systems to support non-disruptive upgrades of hardware and software even during production model training, and for many other reasons.
Small file performance of the storage tier may be critical as many types of inputs, including text, audio, or images will be natively stored as small files. If the storage tier does not handle small files well, an extra step will be required to pre-process and group samples into larger files. Storage, built on top of spinning disks, that relies on SSD as a caching tier, may fall short of the performance needed. Because training with random input batches results in more accurate models, the entire data set must be accessible with full performance. SSD caches only provide high performance for a small subset of the data and will be ineffective at hiding the latency of spinning drives.
Readers will appreciate that the storage systems described above may be configured to support the storage of (among of types of data) blockchains. Such blockchains may be embodied as a continuously growing list of records, called blocks, which are linked and secured using cryptography. Each block in a blockchain may contain a hash pointer as a link to a previous block, a timestamp, transaction data, and so on. Blockchains may be designed to be resistant to modification of the data and can serve as an open, distributed ledger that can record transactions between two parties efficiently and in a verifiable and permanent way. This makes blockchains potentially suitable for the recording of events, medical records, and other records management activities, such as identity management, transaction processing, and others.
Readers will further appreciate that in some embodiments, the storage systems described above may be paired with other resources to support the applications described above. For example, one infrastructure could include primary compute in the form of servers and workstations which specialize in using General-purpose computing on graphics processing units (‘GPGPU’) to accelerate deep learning applications that are interconnected into a computation engine to train parameters for deep neural networks. Each system may have Ethernet external connectivity, InfiniBand external connectivity, some other form of external connectivity, or some combination thereof. In such an example, the GPUs can be grouped for a single large training or used independently to train multiple models. The infrastructure could also include a storage system such as those described above to provide, for example, a scale-out all-flash file or object store through which data can be accessed via high-performance protocols such as NFS, S3, and so on. The infrastructure can also include, for example, redundant top-of-rack Ethernet switches connected to storage and compute via ports in MLAG port channels for redundancy. The infrastructure could also include additional compute in the form of whitebox servers, optionally with GPUs, for data ingestion, pre-processing, and model debugging. Readers will appreciate that additional infrastructures are also be possible.
Readers will appreciate that the systems described above may be better suited for the applications described above relative to other systems that may include, for example, a distributed direct-attached storage (DDAS) solution deployed in server nodes. Such DDAS solutions may be built for handling large, less sequential accesses but may be less able to handle small, random accesses. Readers will further appreciate that the storage systems described above may be utilized to provide a platform for the applications described above that is preferable to the utilization of cloud-based resources as the storage systems may be included in an on-site or in-house infrastructure that is more secure, more locally and internally managed, more robust in feature sets and performance, or otherwise preferable to the utilization of cloud-based resources as part of a platform to support the applications described above. For example, services built on platforms such as IBM's Watson may require a business enterprise to distribute individual user information, such as financial transaction information or identifiable patient records, to other institutions. As such, cloud-based offerings of AI as a service may be less desirable than internally managed and offered AI as a service that is supported by storage systems such as the storage systems described above, for a wide array of technical reasons as well as for various business reasons.
Readers will appreciate that the storage systems described above, either alone or in coordination with other computing machinery may be configured to support other AI related tools. For example, the storage systems may make use of tools like ONXX or other open neural network exchange formats that make it easier to transfer models written in different AI frameworks. Likewise, the storage systems may be configured to support tools like Amazon's Gluon that allow developers to prototype, build, and train deep learning models.”
Readers will further appreciate that the storage systems described above may also be deployed as an edge solution. Such an edge solution may be in place to optimize cloud computing systems by performing data processing at the edge of the network, near the source of the data. Edge computing can push applications, data and computing power (i.e., services) away from centralized points to the logical extremes of a network. Through the use of edge solutions such as the storage systems described above, computational tasks may be performed using the compute resources provided by such storage systems, data may be storage using the storage resources of the storage system, and cloud-based services may be accessed through the use of various resources of the storage system (including networking resources). By performing computational tasks on the edge solution, storing data on the edge solution, and generally making use of the edge solution, the consumption of expensive cloud-based resources may be avoided and, in fact, performance improvements may be experienced relative to a heavier reliance on cloud-based resources.
While many tasks may benefit from the utilization of an edge solution, some particular uses may be especially suited for deployment in such an environment. For example, devices like drones, autonomous cars, robots, and others may require extremely rapid processing—so fast, in fact, that sending data up to a cloud environment and back to receive data processing support may simply be too slow. Likewise, machines like locomotives and gas turbines that generate large amounts of information through the use of a wide array of data-generating sensors may benefit from the rapid data processing capabilities of an edge solution. As an additional example, some IoT devices such as connected video cameras may not be well-suited for the utilization of cloud-based resources as it may be impractical (not only from a privacy perspective, security perspective, or a financial perspective) to send the data to the cloud simply because of the pure volume of data that is involved. As such, many tasks that really on data processing, storage, or communications may be better suited by platforms that include edge solutions such as the storage systems described above.
Consider a specific example of inventory management in a warehouse, distribution center, or similar location. A large inventory, warehousing, shipping, order-fulfillment, manufacturing or other operation has a large amount of inventory on inventory shelves, and high resolution digital cameras that produce a firehose of large data. All of this data may be taken into an image processing system, which may reduce the amount of data to a firehose of small data. All of the small data may be stored on-premises in storage. The on-premises storage, at the edge of the facility, may be coupled to the cloud, for external reports, real-time control and cloud storage. Inventory management may be performed with the results of the image processing, so that inventory can be tracked on the shelves and restocked, moved, shipped, modified with new products, or discontinued/obsolescent products deleted, etc. The above scenario is a prime candidate for an embodiment of the configurable processing and storage systems described above. A combination of compute-only blades and offload blades suited for the image processing, perhaps with deep learning on offload-FPGA or offload-custom blade(s) could take in the firehose of large data from all of the digital cameras, and produce the firehose of small data. All of the small data could then be stored by storage nodes, operating with storage units in whichever combination of types of storage blades best handles the data flow. This is an example of storage and function acceleration and integration. Depending on external communication needs with the cloud, and external processing in the cloud, and depending on reliability of network connections and cloud resources, the system could be sized for storage and compute management with bursty workloads and variable conductivity reliability. Also, depending on other inventory management aspects, the system could be configured for scheduling and resource management in a hybrid edge/cloud environment.
The storage systems described above may also be optimized for use in big data analytics. Big data analytics may be generally described as the process of examining large and varied data sets to uncover hidden patterns, unknown correlations, market trends, customer preferences and other useful information that can help organizations make more-informed business decisions. Big data analytics applications enable data scientists, predictive modelers, statisticians and other analytics professionals to analyze growing volumes of structured transaction data, plus other forms of data that are often left untapped by conventional business intelligence (BI) and analytics programs. As part of that process, semi-structured and unstructured data such as, for example, internet clickstream data, web server logs, social media content, text from customer emails and survey responses, mobile-phone call-detail records, IoT sensor data, and other data may be converted to a structured form. Big data analytics is a form of advanced analytics, which involves complex applications with elements such as predictive models, statistical algorithms and what-if analyses powered by high-performance analytics systems.
The storage systems described above may also support (including implementing as a system interface) applications that perform tasks in response to human speech. For example, the storage systems may support the execution intelligent personal assistant applications such as, for example, Amazon's Alexa, Apple Siri, Google Voice, Samsung Bixby, Microsoft Cortana, and others. While the examples described in the previous sentence make use of voice as input, the storage systems described above may also support chatbots, talkbots, chatterbots, or artificial conversational entities or other applications that are configured to conduct a conversation via auditory or textual methods. Likewise, the storage system may actually execute such an application to enable a user such as a system administrator to interact with the storage system via speech. Such applications are generally capable of voice interaction, music playback, making to-do lists, setting alarms, streaming podcasts, playing audiobooks, and providing weather, traffic, and other real time information, such as news, although in embodiments in accordance with the present disclosure, such applications may be utilized as interfaces to various system management operations.
The storage systems described above may also implement AI platforms for delivering on the vision of self-driving storage. Such AI platforms may be configured to deliver global predictive intelligence by collecting and analyzing large amounts of storage system telemetry data points to enable effortless management, analytics and support. In fact, such storage systems may be capable of predicting both capacity and performance, as well as generating intelligent advice on workload deployment, interaction and optimization. Such AI platforms may be configured to scan all incoming storage system telemetry data against a library of issue fingerprints to predict and resolve incidents in real-time, before they impact customer environments, and captures hundreds of variables related to performance that are used to forecast performance load.
For further explanation, FIG. 4 sets forth a block diagram illustrating a plurality of storage systems (402, 404, 406) that support a pod according to some embodiments of the present disclosure. Although depicted in less detail, the storage systems (402, 404, 406) depicted in FIG. 4 may be similar to the storage systems described above with reference to FIGS. 1A-1D, FIGS. 2A-2G, FIGS. 3A-3B, or any combination thereof. In fact, the storage systems (402, 404, 406) depicted in FIG. 4 may include the same, fewer, or additional components as the storage systems described above.
In the example depicted in FIG. 4, each of the storage systems (402, 404, 406) is depicted as having at least one computer processor (408, 410, 412), computer memory (414, 416, 418), and computer storage (420, 422, 424). Although in some embodiments the computer memory (414, 416, 418) and the computer storage (420, 422, 424) may be part of the same hardware devices, in other embodiments the computer memory (414, 416, 418) and the computer storage (420, 422, 424) may be part of different hardware devices. The distinction between the computer memory (414, 416, 418) and the computer storage (420, 422, 424) in this particular example may be that the computer memory (414, 416, 418) is physically proximate to the computer processors (408, 410, 412) and may store computer program instructions that are executed by the computer processors (408, 410, 412), while the computer storage (420, 422, 424) is embodied as non-volatile storage for storing user data, metadata describing the user data, and so on. Referring to the example above in FIG. 1A, for example, the computer processors (408, 410, 412) and computer memory (414, 416, 418) for a particular storage system (402, 404, 406) may reside within one of more of the controllers (110A-110D) while the attached storage devices (171A-171F) may serve as the computer storage (420, 422, 424) within a particular storage system (402, 404, 406).
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may attach to one or more pods (430, 432) according to some embodiments of the present disclosure. Each of the pods (430, 432) depicted in FIG. 4 can include a dataset (426, 428). For example, a first pod (430) that three storage systems (402, 404, 406) have attached to includes a first dataset (426) while a second pod (432) that two storage systems (404, 406) have attached to includes a second dataset (428). In such an example, when a particular storage system attaches to a pod, the pod's dataset is copied to the particular storage system and then kept up to date as the dataset is modified. Storage systems can be removed from a pod, resulting in the dataset being no longer kept up to date on the removed storage system. In the example depicted in FIG. 4, any storage system which is active for a pod (it is an up-to-date, operating, non-faulted member of a non-faulted pod) can receive and process requests to modify or read the pod's dataset.
In the example depicted in FIG. 4, each pod (430, 432) may also include a set of managed objects and management operations, as well as a set of access operations to modify or read the dataset (426, 428) that is associated with the particular pod (430, 432). In such an example, the management operations may modify or query managed objects equivalently through any of the storage systems. Likewise, access operations to read or modify the dataset may operate equivalently through any of the storage systems. In such an example, while each storage system stores a separate copy of the dataset as a proper subset of the datasets stored and advertised for use by the storage system, the operations to modify managed objects or the dataset performed and completed through any one storage system are reflected in subsequent management objects to query the pod or subsequent access operations to read the dataset.
Readers will appreciate that pods may implement more capabilities than just a clustered synchronously replicated dataset. For example, pods can be used to implement tenants, whereby datasets are in some way securely isolated from each other. Pods can also be used to implement virtual arrays or virtual storage systems where each pod is presented as a unique storage entity on a network (e.g., a Storage Area Network, or Internet Protocol network) with separate addresses. In the case of a multi-storage-system pod implementing a virtual storage system, all physical storage systems associated with the pod may present themselves as in some way the same storage system (e.g., as if the multiple physical storage systems were no different than multiple network ports into a single storage system).
Readers will appreciate that pods may also be units of administration, representing a collection of volumes, file systems, object/analytic stores, snapshots, and other administrative entities, where making administrative changes (e.g., name changes, property changes, managing exports or permissions for some part of the pod's dataset), on any one storage system is automatically reflected to all active storage systems associated with the pod. In addition, pods could also be units of data collection and data analysis, where performance and capacity metrics are presented in ways that aggregate across all active storage systems for the pod, or that call out data collection and analysis separately for each pod, or perhaps presenting each attached storage system's contribution to the incoming content and performance for each a pod.
One model for pod membership may be defined as a list of storage systems, and a subset of that list where storage systems are considered to be in-sync for the pod. A storage system may be considered to be in-sync for a pod if it is at least within a recovery of having identical idle content for the last written copy of the dataset associated with the pod. Idle content is the content after any in-progress modifications have completed with no processing of new modifications. Sometimes this is referred to as “crash recoverable” consistency. Recovery of a pod carries out the process of reconciling differences in applying concurrent updates to in-sync storage systems in the pod. Recovery can resolve any inconsistencies between storage systems in the completion of concurrent modifications that had been requested to various members of the pod but that were not signaled to any requestor as having completed successfully. Storage systems that are listed as pod members but that are not listed as in-sync for the pod can be described as “detached” from the pod. Storage systems that are listed as pod members, are in-sync for the pod, and are currently available for actively serving data for the pod are “online” for the pod.
Each storage system member of a pod may have its own copy of the membership, including which storage systems it last knew were in-sync, and which storage systems it last knew comprised the entire set of pod members. To be online for a pod, a storage system must consider itself to be in-sync for the pod and must be communicating with all other storage systems it considers to be in-sync for the pod. If a storage system can't be certain that it is in-sync and communicating with all other storage systems that are in-sync, then it must stop processing new incoming requests for the pod (or must complete them with an error or exception) until it can be certain that it is in-sync and communicating with all other storage systems that are in-sync. A first storage system may conclude that a second paired storage system should be detached, which will allow the first storage system to continue since it is now in-sync with all storage systems now in the list. But, the second storage system must be prevented from concluding, alternatively, that the first storage system should be detached and with the second storage system continuing operation. This would result in a “split brain” condition that can lead to irreconcilable datasets, dataset corruption, or application corruption, among other dangers.
The situation of needing to determine how to proceed when not communicating with paired storage systems can arise while a storage system is running normally and then notices lost communications, while it is currently recovering from some previous fault, while it is rebooting or resuming from a temporary power loss or recovered communication outage, while it is switching operations from one set of storage system controllers to another set for whatever reason, or during or after any combination of these or other kinds of events. In fact, any time a storage system that is associated with a pod can't communicate with all known non-detached members, the storage system can either wait briefly until communications can be established, go offline and continue waiting, or it can determine through some means that it is safe to detach the non-communicating storage system without risk of incurring a split brain due to the non-communicating storage system concluding the alternative view, and then continue. If a safe detach can happen quickly enough, the storage system can remain online for the pod with little more than a short delay and with no resulting application outages for applications that can issue requests to the remaining online storage systems.
One example of this situation is when a storage system may know that it is out-of-date. That can happen, for example, when a first storage system is first added to a pod that is already associated with one or more storage systems, or when a first storage system reconnects to another storage system and finds that the other storage system had already marked the first storage system as detached. In this case, this first storage system will simply wait until it connects to some other set of storage systems that are in-sync for the pod.
This model demands some degree of consideration for how storage systems are added to or removed from pods or from the in-sync pod members list. Since each storage system will have its own copy of the list, and since two independent storage systems can't update their local copy at exactly the same time, and since the local copy is all that is available on a reboot or in various fault scenarios, care must be taken to ensure that transient inconsistencies don't cause problems. For example, if one storage systems is in-sync for a pod and a second storage system is added, then if the second storage system is updated to list both storage systems as in-sync first, then if there is a fault and a restart of both storage systems, the second might startup and wait to connect to the first storage system while the first might be unaware that it should or could wait for the second storage system. If the second storage system then responds to an inability to connect with the first storage system by going through a process to detach it, then it might succeed in completing a process that the first storage system is unaware of, resulting in a split brain. As such, it may be necessary to ensure that storage systems won't disagree inappropriately on whether they might opt to go through a detach process if they aren't communicating.
One way to ensure that storage systems won't disagree inappropriately on whether they might opt to go through a detach process if they aren't communicating is to ensure that when adding a new storage system to the in-sync member list for a pod, the new storage system first stores that it is a detached member (and perhaps that it is being added as an in-sync member). Then, the existing in-sync storage systems can locally store that the new storage system is an in-sync pod member before the new storage system locally stores that same fact. If there is a set of reboots or network outages prior to the new storage system storing its in-sync status, then the original storage systems may detach the new storage system due to non-communication, but the new storage system will wait. A reverse version of this change might be needed for removing a communicating storage system from a pod: first the storage system being removed stores that it is no longer in-sync, then the storage systems that will remain store that the storage system being removed is no longer in-sync, then all storage systems delete the storage system being removed from their pod membership lists. Depending on the implementation, an intermediate persisted detached state may not be necessary. Whether or not care is required in local copies of membership lists may depend on the model storage systems use for monitoring each other or for validating their membership. If a consensus model is used for both, or if an external system (or an external distributed or clustered system) is used to store and validate pod membership, then inconsistencies in locally stored membership lists may not matter.
When communications fail or one or several storage systems in a pod fail, or when a storage system starts up (or fails over to a secondary controller) and can't communicate with paired storage systems for a pod, and it is time for one or more storage systems to decide to detach one or more paired storage systems, some algorithm or mechanism must be employed to decide that it is safe to do so and to follow through on the detach. One means of resolving detaches is use a majority (or quorum) model for membership. With three storage systems, as long as two are communicating, they can agree to detach a third storage system that isn't communicating, but that third storage system cannot by itself choose to detach either of the other two. Confusion can arise when storage system communication is inconsistent. For example, storage system A might be communicating with storage system B but not C, while storage system B might be communicating with both A and C. So, A and B could detach C, or B and C could detach A, but more communication between pod members may be needed to figure this out.
Care needs to be taken in a quorum membership model when adding and removing storage systems. For example, if a fourth storage system is added, then a “majority” of storage systems is at that point three. The transition from three storage systems (with two required for majority) to a pod including a fourth storage system (with three required for majority) may require something similar to the model described previously for carefully adding a storage system to the in-sync list. For example, the fourth storage system might start in an attaching state but not yet attached where it would never instigate a vote over quorum. Once in that state, the original three pod members could each be updated to be aware of the fourth member and the new requirement for a three storage system majority to detach a fourth. Removing a storage system from a pod might similarly move that storage system to a locally stored “detaching” state before updating other pod members. A variant scheme for this is to use a distributed consensus mechanism such as PAXOS or RAFT to implement any membership changes or to process detach requests.
Another means of managing membership transitions is to use an external system that is outside of the storage systems themselves to handle pod membership. In order to become online for a pod, a storage system must first contact the external pod membership system to verify that it is in-sync for the pod. Any storage system that is online for a pod should then remain in communication with the pod membership system and should wait or go offline if it loses communication. An external pod membership manager could be implemented as a highly available cluster using various cluster tools, such as Oracle RAC, Linux HA, VERITAS Cluster Server, IBM's HACMP, or others. An external pod membership manager could also use distributed configuration tools such as Etcd or Zookeeper, or a reliable distributed database such as Amazon's DynamoDB.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may receive a request to read a portion of the dataset (426, 428) and process the request to read the portion of the dataset locally according to some embodiments of the present disclosure. Readers will appreciate that although requests to modify (e.g., a write operation) the dataset (426, 428) require coordination between the storage systems (402, 404, 406) in a pod, as the dataset (426, 428) should be consistent across all storage systems (402, 404, 406) in a pod, responding to a request to read a portion of the dataset (426, 428) does not require similar coordination between the storage systems (402, 404, 406). As such, a particular storage system that receives a read request may service the read request locally by reading a portion of the dataset (426, 428) that is stored within the storage system's storage devices, with no synchronous communication with other storage systems in the pod. Read requests received by one storage system for a replicated dataset in a replicated cluster are expected to avoid any communication in the vast majority of cases, at least when received by a storage system that is running within a cluster that is also running nominally. Such reads should normally be processed simply by reading from the local copy of a clustered dataset with no further interaction required with other storage systems in the cluster
Readers will appreciate that the storage systems may take steps to ensure read consistency such that a read request will return the same result regardless of which storage system processes the read request. For example, the resulting clustered dataset content for any set of updates received by any set of storage systems in the cluster should be consistent across the cluster, at least at any time updates are idle (all previous modifying operations have been indicated as complete and no new update requests have been received and processed in any way). More specifically, the instances of a clustered dataset across a set of storage systems can differ only as a result of updates that have not yet completed. This means, for example, that any two write requests which overlap in their volume block range, or any combination of a write request and an overlapping snapshot, compare-and-write, or virtual block range copy, must yield a consistent result on all copies of the dataset. Two operations should not yield a result as if they happened in one order on one storage system and a different order on another storage system in the replicated cluster.
Furthermore, read requests can be made time order consistent. For example, if one read request is received on a replicated cluster and completed and that read is then followed by another read request to an overlapping address range which is received by the replicated cluster and where one or both reads in any way overlap in time and volume address range with a modification request received by the replicated cluster (whether any of the reads or the modification are received by the same storage system or a different storage system in the replicated cluster), then if the first read reflects the result of the update then the second read should also reflect the results of that update, rather than possibly returning data that preceded the update. If the first read does not reflect the update, then the second read can either reflect the update or not. This ensures that between two read requests “time” for a data segment cannot roll backward.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also detect a disruption in data communications with one or more of the other storage systems and determine whether to the particular storage system should remain in the pod. A disruption in data communications with one or more of the other storage systems may occur for a variety of reasons. For example, a disruption in data communications with one or more of the other storage systems may occur because one of the storage systems has failed, because a network interconnect has failed, or for some other reason. An important aspect of synchronous replicated clustering is ensuring that any fault handling doesn't result in unrecoverable inconsistencies, or any inconsistency in responses. For example, if a network fails between two storage systems, at most one of the storage systems can continue processing newly incoming I/O requests for a pod. And, if one storage system continues processing, the other storage system can't process any new requests to completion, including read requests.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also determine whether the particular storage system should remain in the pod in response to detecting a disruption in data communications with one or more of the other storage systems. As mentioned above, to be ‘online’ as part of a pod, a storage system must consider itself to be in-sync for the pod and must be communicating with all other storage systems it considers to be in-sync for the pod. If a storage system can't be certain that it is in-sync and communicating with all other storage systems that are in-sync, then it may stop processing new incoming requests to access the dataset (426, 428). As such, the storage system may determine whether the particular storage system should remain online as part of the pod, for example, by determining whether it can communicate with all other storage systems it considers to be in-sync for the pod (e.g., via one or more test messages), by determining whether all other storage systems it considers to be in-sync for the pod also consider the storage system to be attached to the pod, through a combination of both steps where the particular storage system must confirm that it can communicate with all other storage systems it considers to be in-sync for the pod and that all other storage systems it considers to be in-sync for the pod also consider the storage system to be attached to the pod, or through some other mechanism.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also keep the dataset on the particular storage system accessible for management and dataset operations in response to determining that the particular storage system should remain in the pod. The storage system may keep the dataset (426, 428) on the particular storage system accessible for management and dataset operations, for example, by accepting requests to access the version of the dataset (426, 428) that is stored on the storage system and processing such requests, by accepting and processing management operations associated with the dataset (426, 428) that are issued by a host or authorized administrator, by accepting and processing management operations associated with the dataset (426, 428) that are issued by one of the other storage systems, or in some other way.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may, however, make the dataset on the particular storage system inaccessible for management and dataset operations in response to determining that the particular storage system should not remain in the pod. The storage system may make the dataset (426, 428) on the particular storage system inaccessible for management and dataset operations, for example, by rejecting requests to access the version of the dataset (426, 428) that is stored on the storage system, by rejecting management operations associated with the dataset (426, 428) that are issued by a host or other authorized administrator, by rejecting management operations associated with the dataset (426, 428) that are issued by one of the other storage systems in the pod, or in some other way.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also detect that the disruption in data communications with one or more of the other storage systems has been repaired and make the dataset on the particular storage system accessible for management and dataset operations. The storage system may detect that the disruption in data communications with one or more of the other storage systems has been repaired, for example, by receiving a message from the one or more of the other storage systems. In response to detecting that the disruption in data communications with one or more of the other storage systems has been repaired, the storage system may make the dataset (426, 428) on the particular storage system accessible for management and dataset operations once the previously detached storage system has been resynchronized with the storage systems that remained attached to the pod.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also go offline from the pod such that the particular storage system no longer allows management and dataset operations. The depicted storage systems (402, 404, 406) may go offline from the pod such that the particular storage system no longer allows management and dataset operations for a variety of reasons. For example, the depicted storage systems (402, 404, 406) may also go offline from the pod due to some fault with the storage system itself, because an update or some other maintenance is occurring on the storage system, due to communications faults, or for many other reasons. In such an example, the depicted storage systems (402, 404, 406) may subsequently update the dataset on the particular storage system to include all updates to the dataset since the particular storage system went offline and go back online with the pod such that the particular storage system allows management and dataset operations, as will be described in greater detail in the resynchronization sections included below.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also identifying a target storage system for asynchronously receiving the dataset, where the target storage system is not one of the plurality of storage systems across which the dataset is synchronously replicated. Such a target storage system may represent, for example, a backup storage system, as some storage system that makes use of the synchronously replicated dataset, and so on. In fact, synchronous replication can be leveraged to distribute copies of a dataset closer to some rack of servers, for better local read performance. One such case is smaller top-of-rack storage systems symmetrically replicated to larger storage systems that are centrally located in the data center or campus and where those larger storage systems are more carefully managed for reliability or are connected to external networks for asynchronous replication or backup services.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also identify a portion of the dataset that is not being asynchronously replicated to the target storage system by any of the other storages systems and asynchronously replicate, to the target storage system, the portion of the dataset that is not being asynchronously replicated to the target storage system by any of the other storages systems, wherein the two or more storage systems collectively replicate the entire dataset to the target storage system. In such a way, the work associated with asynchronously replicating a particular dataset may be split amongst the members of a pod, such that each storage system in a pod is only responsible for asynchronously replicating a subset of a dataset to the target storage system.
In the example depicted in FIG. 4, the depicted storage systems (402, 404, 406) may also detach from the pod, such that the particular storage system that detaches from the pod is no longer included in the set of storage systems across which the dataset is synchronously replicated. For example, if storage system (404) in FIG. 4 detached from the pod (430) illustrated in FIG. 4, the pod (430) would only include storage systems (402, 406) as the storage systems across which the dataset (426) that is included in the pod (430) would be synchronously replicated across. In such an example, detaching the storage system from the pod could also include removing the dataset from the particular storage system that detached from the pod. Continuing with the example where the storage system (404) in FIG. 4 detached from the pod (430) illustrated in FIG. 4, the dataset (426) that is included in the pod (430) could be deleted or otherwise removed from the storage system (404).
Readers will appreciate that there are a number of unique administrative capabilities enabled by the pod model that can further be supported. Also, the pod model itself introduces some issues that can be addressed by an implementation. For example, when a storage system is offline for a pod, but is otherwise running, such as because an interconnect failed and another storage system for the pod won out in mediation, there may still be a desire or need to access the offline pod's dataset on the offline storage system. One solution may be simply to enable the pod in some detached mode and allow the dataset to be accessed. However, that solution can be dangerous and that solution can cause the pod's metadata and data to be much more difficult to reconcile when the storage systems do regain communication. Furthermore, there could still be a separate path for hosts to access the offline storage system as well as the still online storage systems. In that case, a host might issue I/O to both storage systems even though they are no longer being kept in sync, because the host sees target ports reporting volumes with the same identifiers and the host I/O drivers presume it sees additional paths to the same volume. This can result in fairly damaging data corruption as reads and writes issued to both storage systems are no longer consistent even though the host presumes they are. As a variant of this case, in a clustered application, such as a shared storage clustered database, the clustered application running on one host might be reading or writing to one storage system and the same clustered application running on another host might be reading or writing to the “detached” storage system, yet the two instances of the clustered application are communicating between each other on the presumption that the dataset they each see is entirely consistent for completed writes. Since they aren't consistent, that presumption is violated and the application's dataset (e.g., the database) can quickly end up being corrupted.
One way to solve both of these problems is to allow for an offline pod, or perhaps a snapshot of an offline pod, to be copied to a new pod with new volumes that have sufficiently new identities that host I/O drivers and clustered applications won't confuse the copied volumes as being the same as the still online volumes on another storage system. Since each pod maintains a complete copy of the dataset, which is crash consistent but perhaps slightly different from the copy of the pod dataset on another storage system, and since each pod has an independent copy of all data and metadata needed to operate on the pod content, it is a straightforward problem to make a virtual copy of some or all volumes or snapshots in the pod to new volumes in a new pod. In a logical extent graph implementation, for example, all that is needed is to define new volumes in a new pod which reference logical extent graphs from the copied pod associated with the pod's volumes or snapshots, and with the logical extent graphs being marked as copy on write. The new volumes should be treated as new volumes, similarly to how volume snapshots copied to a new volume might be implemented. Volumes may have the same administrative name, though within a new pod namespace. But, they should have different underlying identifiers, and differing logical unit identifiers from the original volumes.
In some cases it may be possible to use virtual network isolation techniques (for example, by creating a virtual LAN in the case of IP networks or a virtual SAN in the case of fiber channel networks) in such a way that isolation of volumes presented to some interfaces can be assured to be inaccessible from host network interfaces or host SCSI initiator ports that might also see the original volumes. In such cases, it may be safe to provide the copies of volumes with the same SCSI or other storage identifiers as the original volumes. This could be used, for example, in cases where the applications expect to see a particular set of storage identifiers in order to function without an undue burden in reconfiguration.
Some of the techniques described herein could also be used outside of an active fault context to test readiness for handling faults. Readiness testing (sometimes referred to as “fire drills”) is commonly required for disaster recovery configurations, where frequent and repeated testing is considered a necessity to ensure that most or all aspects of a disaster recovery plan are correct and account for any recent changes to applications, datasets, or changes in equipment. Readiness testing should be non-disruptive to current production operations, including replication. In many cases the real operations can't actually be invoked on the active configuration, but a good way to get close is to use storage operations to make copies of production datasets, and then perhaps couple that with the use of virtual networking, to create an isolated environment containing all data that is believed necessary for the important applications that must be brought up successfully in cases of disasters. Making such a copy of a synchronously replicated (or even an asynchronously replicated) dataset available within a site (or collection of sites) that is expected to perform a disaster recovery readiness test procedure and then starting the important applications on that dataset to ensure that it can startup and function is a great tool, since it helps ensure that no important parts of the application datasets were left out in the disaster recovery plan. If necessary, and practical, this could be coupled with virtual isolated networks coupled perhaps with isolated collection of physical or virtual machines, to get as close as possible to a real world disaster recovery takeover scenario. Virtually copying a pod (or set of pods) to another pod as a point-in-time image of the pod datasets immediately creates an isolated dataset that contains all the copied elements and that can then be operated on essentially identically to the originally pods, as well as allowing isolation to a single site (or a few sites) separately from the original pod. Further, these are fast operations and they can be torn down and repeated easily allowing testing to repeated as often as is desired.
Some enhancements could be made to get further toward perfect disaster recovery testing. For example, in conjunction with isolated networks, SCSI logical unit identities or other types of identities could be copied into the target pod so that the test servers, virtual machines, and applications see the same identities. Further, the administrative environment of the servers could be configured to respond to requests from a particular virtual set of virtual networks to respond to requests and operations on the original pod name so scripts don't require use of test-variants with alternate “test” versions of object names. A further enhancement can be used in cases where the host-side server infrastructure that will take over in the case of a disaster takeover can be used during a test. This includes cases where a disaster recovery data center is completely stocked with alternative server infrastructure that won't generally be used until directed to do so by a disaster. It also includes cases where that infrastructure might be used for non-critical operations (for example, running analytics on production data, or simply supporting application development or other functions which may be important but can be halted if needed for more critical functions). Specifically, host definitions and configurations and the server infrastructure that will use them can be set up as they will be for an actual disaster recovery takeover event and tested as part of disaster recovery takeover testing, with the tested volumes being connected to these host definitions from the virtual pod copy used to provide a snapshot of the dataset. From the standpoint of the storage systems involved, then, these host definitions and configurations used for testing, and the volume-to-host connection configurations used during testing, can be reused when an actual disaster takeover event is triggered, greatly minimizing the configuration differences between the test configuration and the real configuration that will be used in case of a disaster recovery takeover.
In some cases it may make sense to move volumes out of a first pod and into a new second pod including just those volumes. The pod membership and high availability and recovery characteristics can then be adjusted separately, and administration of the two resulting pod datasets can then be isolated from each other. An operation that can be done in one direction should also be possible in the other direction. At some point, it may make sense to take two pods and merge them into one so that the volumes in each of the original two pods will now track each other for storage system membership and high availability and recovery characteristics and events. Both operations can be accomplished safely and with reasonably minimal or no disruption to running applications by relying on the characteristics suggested for changing mediation or quorum properties for a pod which were discussed in an earlier section. With mediation, for example, a mediator for a pod can be changed using a sequence consisting of a step where each storage system in a pod is changed to depend on both a first mediator and a second mediator and each is then changed to depend only on the second mediator. If a fault occurs in the middle of the sequence, some storage systems may depend on both the first mediator and the second mediator, but in no case will recovery and fault handling result in some storage systems depending only on the first mediator and other storage systems only depending on the second mediator. Quorum can be handled similarly by temporarily depending on winning against both a first quorum model and a second quorum model in order to proceed to recovery. This may result in a very short time period where availability of the pod in the face of faults depend on additional resources, thus reducing potential availability, but this time period is very short and the reduction in availability is often very little. With mediation, if the change in mediator parameters is nothing more than the change in the key used for mediation and the mediation service used is the same, then the potential reduction in availability is even less, since it now depends only on two calls to the same service versus one call to that service, and rather than separate calls to two separate services.
Readers will note that changing the quorum model may be quite complex. An additional step may be necessary where storage systems will participate in the second quorum model but won't depend on winning in that second quorum model, which is then followed by the step of also depending on the second quorum model. This may be necessary to account for the fact that if only one system has processed the change to depend on the quorum model, then it will never win quorum since there will never be a majority. With this model in place for changing the high availability parameters (mediation relationship, quorum model, takeover preferences), we can create a safe procedure for these operations to split a pod into two or to join two pods into one. This may require adding one other capability: linking a second pod to a first pod for high availability such that if two pods include compatible high availability parameters the second pod linked to the first pod can depend on the first pod for determining and instigating detach-related processing and operations, offline and in-sync states, and recovery and resynchronization actions.
To split a pod into two, which is an operation to move some volumes into a newly created pod, a distributed operation may be formed that can be described as: form a second pod into which we will move a set of volumes which were previously in a first pod, copy the high availability parameters from the first pod into the second pod to ensure they are compatible for linking, and link the second pod to the first pod for high availability. This operation may be encoded as messages and should be implemented by each storage system in the pod in such a way that the storage system ensures that the operation happens completely on that storage system or does not happen at all if processing is interrupted by a fault. Once all in-sync storage systems for the two pods have processed this operation, the storage systems can then process a subsequent operation which changes the second pod so that it is no longer linked to the first pod. As with other changes to high availability characteristics for a pod, this involves first having each in-sync storage system change to rely on both the previous model (that model being that high availability is linked to the first pod) and the new model (that model being its own now independent high availability). In the case of mediation or quorum, this means that storage systems which processed this change will first depend on mediation or quorum being achieved as appropriate for the first pod and will additionally depend on a new separate mediation (for example, a new mediation key) or quorum being achieved for the second pod before the second pod can proceed following a fault that required mediation or testing for quorum. As with the previous description of changing quorum models, an intermediate step may set storage systems to participate in quorum for the second pod before the step where storage systems participate in and depend on quorum for the second pod. Once all in-sync storage systems have processed the change to depend on the new parameters for mediation or quorum for both the first pod and the second pod, the split is complete.
Joining a second pod into a first pod operates essentially in reverse. First, the second pod must be adjusted to be compatible with the first pod, by having an identical list of storage systems and by having a compatible high availability model. This may involve some set of steps such as those described elsewhere in this paper to add or remove storage systems or to change mediator and quorum models. Depending on implementation, it may be necessary only to reach an identical list of storage systems. Joining proceeds by processing an operation on each in-sync storage system to link the second pod to the first pod for high availability. Each storage system which processes that operation will then depend on the first pod for high availability and then the second pod for high availability. Once all in-sync storage systems for the second pod have processed that operation, the storage systems will then each process a subsequent operation to eliminate the link between the second pod and the first pod, migrate the volumes from the second pod into the first pod, and delete the second pod. Host or application dataset access can be preserved throughout these operations, as long as the implementation allows proper direction of host or application dataset modification or read operations to the volume by identity and as long as the identity is preserved as appropriate to the storage protocol or storage model (for example, as long as logical unit identifiers for volumes and use of target ports for accessing volumes are preserved in the case of SCSI).
Migrating a volume between pods may present issues. If the pods have an identical set of in-sync membership storage systems, then it may be straightforward: temporarily suspend operations on the volumes being migrated, switch control over operations on those volumes to controlling software and structures for the new pod, and then resume operations. This allows for a seamless migration with continuous uptime for applications apart from the very brief operation suspension, provided network and ports migrate properly between pods. Depending on the implementation, suspending operations may not even be necessary, or may be so internal to the system that the suspension of operations has no impact. Copying volumes between pods with different in-sync membership sets is more of a problem. If the target pod for the copy has a subset of in-sync members from the source pod, this isn't much of a problem: a member storage system can be dropped safely enough without having to do more work. But, if the target pod adds in-sync member storage systems to the volume over the source pod, then the added storage systems must be synchronized to include the volume's content before they can be used. Until synchronized, this leaves the copied volumes distinctly different from the already synchronized volumes, in that fault handling differs and request handling from the not yet synced member storage systems either won't work or must be forwarded or won't be as fast because reads will have to traverse an interconnect. Also, the internal implementation will have to handle some volumes being in sync and ready for fault handling and others not being in sync.
There are other problems relating to reliability of the operation in the face of faults. Coordinating a migration of volumes between multi-storage-system pods is a distributed operation. If pods are the unit of fault handling and recovery, and if mediation or quorum or whatever means are used to avoid split-brain situations, then a switch in volumes from one pod with a particular set of state and configurations and relationships for fault handling, recovery, mediation and quorum to another then storage systems in a pod have to be careful about coordinating changes related to that handling for any volumes. Operations can't be atomically distributed between storage systems, but must be staged in some way. Mediation and quorum models essentially provide pods with the tools for implementing distributed transactional atomicity, but this may not extend to inter-pod operations without adding to the implementation.
Consider even a simple migration of a volume from a first pod to a second pod even for two pods that share the same first and second storage systems. At some point the storage systems will coordinate to define that the volume is now in the second pod and is no longer in the first pod. If there is no inherent mechanism for transactional atomicity across the storage systems for the two pods, then a naive implementation could leave the volume in the first pod on the first storage system and the second pod on the second storage system at the time of a network fault that results in fault handling to detach storage systems from the two pods. If pods separately determine which storage system succeeds in detaching the other, then the result could be that the same storage system detaches the other storage system for both pods, in which case the result of the volume migration recovery should be consistent, or it could result in a different storage system detaching the other for the two pods. If the first storage system detaches the second storage system for the first pod and the second storage system detaches the first storage system for the second pod, then recovery might result in the volume being recovered to the first pod on the first storage system and into the second pod on the second storage system, with the volume then running and exported to hosts and storage applications on both storage systems. If instead the second storage system detaches the first storage system for the first pod and first storage detaches the second storage system for the second pod, then recovery might result in the volume being discarded from the second pod by the first storage system and the volume being discarded from the first pod by the second storage system, resulting in the volume disappearing entirely. If the pods a volume is being migrated between are on differing sets of storage systems, then things can get even more complicated.
A solution to these problems may be to use an intermediate pod along with the techniques described previously for splitting and joining pods. This intermediate pod may never be presented as visible managed objects associated with the storage systems. In this model, volumes to be moved from a first pod to a second pod are first split from the first pod into a new intermediate pod using the split operation described previously. The storage system members for the intermediate pod can then be adjusted to match the membership of storage systems by adding or removing storage systems from the pod as necessary. Subsequently, the intermediate pod can be joined with the second pod.
For further explanation, FIG. 5 sets forth a flow chart illustrating steps that may be performed by storage systems (402, 404, 406) that support a pod according to some embodiments of the present disclosure. Although depicted in less detail, the storage systems (402. 404, 406) depicted in FIG. 5 may be similar to the storage systems described above with reference to FIGS. 1A-1D, FIGS. 2A-2G, FIGS. 3A-3B, FIG. 4, or any combination thereof. In fact, the storage systems (402, 404, 406) depicted in FIG. 5 may include the same, fewer, additional components as the storage systems described above.
In the example method depicted in FIG. 5, a storage system (402) may attach (508) to a pod. The model for pod membership may include a list of storage systems and a subset of that list where storage systems are presumed to be in-sync for the pod. A storage system is in-sync for a pod if it is at least within a recovery of having identical idle content for the last written copy of the dataset associated with the pod. Idle content is the content after any in-progress modifications have completed with no processing of new modifications. Sometimes this is referred to as “crash recoverable” consistency. Storage systems that are listed as pod members but that are not listed as in-sync for the pod can be described as “detached” from the pod. Storage systems that are listed as pod members, are in-sync for the pod, and are currently available for actively serving data for the pod are “online” for the pod.
In the example method depicted in FIG. 5, the storage system (402) may attach (508) to a pod, for example, by synchronizing its locally stored version of the dataset (426) along with an up-to-date version of the dataset (426) that is stored on other storage systems (404, 406) in the pod that are online, as the term is described above. In such an example, in order for the storage system (402) to attach (508) to the pod, a pod definition stored locally within each of the storage systems (402, 404, 406) in the pod may need to be updated in order for the storage system (402) to attach (508) to the pod. In such an example, each storage system member of a pod may have its own copy of the membership, including which storage systems it last knew were in-sync, and which storage systems it last knew comprised the entire set of pod members.
In the example method depicted in FIG. 5, the storage system (402) may also receive (510) a request to read a portion of the dataset (426) and the storage system (402) may process (512) the request to read the portion of the dataset (426) locally. Readers will appreciate that although requests to modify (e.g., a write operation) the dataset (426) require coordination between the storage systems (402, 404, 406) in a pod, as the dataset (426) should be consistent across all storage systems (402, 404, 406) in a pod, responding to a request to read a portion of the dataset (426) does not require similar coordination between the storage systems (402, 404, 406). As such, a particular storage system (402) that receives a read request may service the read request locally by reading a portion of the dataset (426) that is stored within the storage system's (402) storage devices, with no synchronous communication with other storage systems (404, 406) in the pod. Read requests received by one storage system for a replicated dataset in a replicated cluster are expected to avoid any communication in the vast majority of cases, at least when received by a storage system that is running within a cluster that is also running nominally. Such reads should normally be processed simply by reading from the local copy of a clustered dataset with no further interaction required with other storage systems in the cluster
Readers will appreciate that the storage systems may take steps to ensure read consistency such that a read request will return the same result regardless of which storage system processes the read request. For example, the resulting clustered dataset content for any set of updates received by any set of storage systems in the cluster should be consistent across the cluster, at least at any time updates are idle (all previous modifying operations have been indicated as complete and no new update requests have been received and processed in any way). More specifically, the instances of a clustered dataset across a set of storage systems can differ only as a result of updates that have not yet completed. This means, for example, that any two write requests which overlap in their volume block range, or any combination of a write request and an overlapping snapshot, compare-and-write, or virtual block range copy, must yield a consistent result on all copies of the dataset. Two operations cannot yield a result as if they happened in one order on one storage system and a different order on another storage system in the replicated cluster.
Furthermore, read requests may be time order consistent. For example, if one read request is received on a replicated cluster and completed and that read is then followed by another read request to an overlapping address range which is received by the replicated cluster and where one or both reads in any way overlap in time and volume address range with a modification request received by the replicated cluster (whether any of the reads or the modification are received by the same storage system or a different storage system in the replicated cluster), then if the first read reflects the result of the update then the second read should also reflect the results of that update, rather than possibly returning data that preceded the update. If the first read does not reflect the update, then the second read can either reflect the update or not. This ensures that between two read requests “time” for a data segment cannot roll backward.
In the example method depicted in FIG. 5, the storage system (402) may also detect (514) a disruption in data communications with one or more of the other storage systems (404, 406). A disruption in data communications with one or more of the other storage systems (404, 406) may occur for a variety of reasons. For example, a disruption in data communications with one or more of the other storage systems (404, 406) may occur because one of the storage systems (402, 404, 406) has failed, because a network interconnect has failed, or for some other reason. An important aspect of synchronous replicated clustering is ensuring that any fault handling doesn't result in unrecoverable inconsistencies, or any inconsistency in responses. For example, if a network fails between two storage systems, at most one of the storage systems can continue processing newly incoming I/O requests for a pod. And, if one storage system continues processing, the other storage system can't process any new requests to completion, including read requests.
In the example method depicted in FIG. 5, the storage system (402) may also determine (516) whether to the particular storage system (402) should remain online as part of the pod. As mentioned above, to be ‘online’ as part of a pod, a storage system must consider itself to be in-sync for the pod and must be communicating with all other storage systems it considers to be in-sync for the pod. If a storage system can't be certain that it is in-sync and communicating with all other storage systems that are in-sync, then it may stop processing new incoming requests to access the dataset (426). As such, the storage system (402) may determine (516) whether to the particular storage system (402) should remain online as part of the pod, for example, by determining whether it can communicate with all other storage systems (404, 406) it considers to be in-sync for the pod (e.g., via one or more test messages), by determining whether the all other storage systems (404, 406) it considers to be in-sync for the pod also consider the storage system (402) to be attached to the pod, through a combination of both steps where the particular storage system (402) must confirm that it can communicate with all other storage systems (404, 406) it considers to be in-sync for the pod and that all other storage systems (404, 406) it considers to be in-sync for the pod also consider the storage system (402) to be attached to the pod, or through some other mechanism.
In the example method depicted in FIG. 5, the storage system (402) may also, responsive to affirmatively (518) determining that the particular storage system (402) should remain online as part of the pod, keep (522) the dataset (426) on the particular storage system (402) accessible for management and dataset operations. The storage system (402) may keep (522) the dataset (426) on the particular storage system (402) accessible for management and dataset operations, for example, by accepting requests to access the version of the dataset (426) that is stored on the storage system (402) and processing such requests, by accepting and processing management operations associated with the dataset (426) that are issued by a host or authorized administrator, by accepting and processing management operations associated with the dataset (426) that are issued by one of the other storage systems (404, 406) in the pod, or in some other way.
In the example method depicted in FIG. 5, the storage system (402) may also, responsive to determining that the particular storage system should not (520) remain online as part of the pod, make (524) the dataset (426) on the particular storage system (402) inaccessible for management and dataset operations. The storage system (402) may make (524) the dataset (426) on the particular storage system (402) inaccessible for management and dataset operations, for example, by rejecting requests to access the version of the dataset (426) that is stored on the storage system (402), by rejecting management operations associated with the dataset (426) that are issued by a host or other authorized administrator, by rejecting management operations associated with the dataset (426) that are issued by one of the other storage systems (404, 406) in the pod, or in some other way.
In the example method depicted in FIG. 5, the storage system (402) may also detect (526) that the disruption in data communications with one or more of the other storage systems (404, 406) has been repaired. The storage system (402) may detect (526) that the disruption in data communications with one or more of the other storage systems (404, 406) has been repaired, for example, by receiving a message from the one or more of the other storage systems (404, 406). In response to detecting (526) that the disruption in data communications with one or more of the other storage systems (404, 406) has been repaired, the storage system (402) may make (528) the dataset (426) on the particular storage system (402) accessible for management and dataset operations.
Readers will appreciate that the example depicted in FIG. 5 describes an embodiment in which various actions are depicted as occurring within some order, although no ordering is required. Furthermore, other embodiments may exist where the storage system (402) only carries out a subset of the described actions. For example, the storage system (402) may perform the steps of detecting (514) a disruption in data communications with one or more of the other storage systems (404, 406), determining (516) whether to the particular storage system (402) should remain in the pod, keeping (522) the dataset (426) on the particular storage system (402) accessible for management and dataset operations or making (524) the dataset (426) on the particular storage system (402) inaccessible for management and dataset operations without first receiving (510) a request to read a portion of the dataset (426) and processing (512) the request to read the portion of the dataset (426) locally. Furthermore, the storage system (402) may detect (526) that the disruption in data communications with one or more of the other storage systems (404, 406) has been repaired and make (528) the dataset (426) on the particular storage system (402) accessible for management and dataset operations without first receiving (510) a request to read a portion of the dataset (426) and processing (512) the request to read the portion of the dataset (426) locally. In fact, none of the steps described herein are explicitly required in all embodiments as prerequisites for performing other steps described herein.
For further explanation, FIG. 6 sets forth a flow chart illustrating steps that may be performed by storage systems (402, 404, 406) that support a pod according to some embodiments of the present disclosure. Although depicted in less detail, the storage systems (402. 404, 406) depicted in FIG. 6 may be similar to the storage systems described above with reference to FIGS. 1A-1D, FIGS. 2A-2G, FIGS. 3A-3B, FIG. 4, or any combination thereof. In fact, the storage systems (402, 404, 406) depicted in FIG. 6 may include the same, fewer, additional components as the storage systems described above.
In the example method depicted in FIG. 6, two or more of the storage systems (402, 404) may each identify (608) a target storage system (618) for asynchronously receiving the dataset (426). The target storage system (618) for asynchronously receiving the dataset (426) may be embodied, for example, as a backup storage system that is located in a different data center than either of the storage systems (402, 404) that are members of a particular pod, as cloud storage that is provided by a cloud services provider, or in many other ways. Readers will appreciate that the target storage system (618) is not one of the plurality of storage systems (402, 404) across which the dataset (426) is synchronously replicated, and as such, the target storage system (618) initially does not include an up-to-date local copy of the dataset (426).
In the example method depicted in FIG. 6, two or more of the storage systems (402, 404) may each also identify (610) a portion of the dataset (426) that is not being asynchronously replicated to the target storage (618) system by any of the other storages systems that are members of a pod that includes the dataset (426). In such an example, the storage systems (402, 404) may each asynchronously replicate (612), to the target storage system (618), the portion of the dataset (426) that is not being asynchronously replicated to the target storage system by any of the other storages systems. Consider an example in which a first storage system (402) is responsible for asynchronously replicating a first portion (e.g., a first half of an address space) of the dataset (426) to the target storage system (618). In such an example, the second storage system (404) would be responsible for asynchronously replicating a second portion (e.g., a second half of an address space) of the dataset (426) to the target storage system (618), such that the two or more storage systems (402, 404) collectively replicate the entire dataset (426) to the target storage system (618).
Readers will appreciate that through the use of pods, as described above, the replication relationship between two storage systems may be switched from a relationship where data is asynchronously replicated to a relationship where data is synchronously replicated. For example, if storage system A is configured to asynchronously replicate a dataset to storage system B, creating a pod that includes the dataset, storage system A as a member, and storage system B as a member can switch the relationship where data is asynchronously replicated to a relationship where data is synchronously replicated. Likewise, through the use of pods, the replication relationship between two storage systems may be switched from a relationship where data is synchronously replicated to a relationship where data is asynchronously replicated. For example, if a pod is created that includes the dataset, storage system A as a member, and storage system B as a member, by merely unstretching the pod (to remove storage system A as a member or to remove storage system B as a member), a relationship where data is synchronously replicated between the storage systems can immediately be switched to a relationship where data is asynchronously replicated. In such a way, storage systems may switch back-and-forth as needed between asynchronous replication and synchronous replication.
This switching can be facilitated by the implementation relying on similar techniques for both synchronous and asynchronous replication. For example, if resynchronization for a synchronously replicated dataset relies on the same or a compatible mechanism as is used for asynchronous replication, then switching to asynchronous replication is conceptually identical to dropping the in-sync state and leaving a relationship in a state similar to a “perpetual recovery” mode. Likewise, switching from asynchronous replication to synchronous replication can operate conceptually by “catching up” and becoming in-sync just as is done when completing a resynchronization with the switching system becoming an in-sync pod member.
Alternatively, or additionally, if both synchronous and asynchronous replication rely on similar or identical common metadata, or a common model for representing and identifying logical extents or stored block identities, or a common model for representing content-addressable stored blocks, then these aspects of commonality can be leveraged to dramatically reduce the content that may need to be transferred when switching to and from synchronous and asynchronous replication. Further, if a dataset is asynchronously replicated from a storage system A to a storage system B, and system B further asynchronously replicates that data set to a storage system C, then a common metadata model, common logical extent or block identities, or common representation of content-addressable stored blocks, can dramatically reduce the data transfers needed to enable synchronous replication between storage system A and storage system C.
Readers will further appreciate that that through the use of pods, as described above, replication techniques may be used to perform tasks other than replicating data. In fact, because a pod may include a set of managed objects, tasks like migrating a virtual machine may be carried out using pods and the replication techniques described herein. For example, if virtual machine A is executing on storage system A, by creating a pod that includes virtual machine A as a managed object, storage system A as a member, and storage system B as a member, virtual machine A and any associated images and definitions may be migrated to storage system B, at which time the pod could simply be destroyed, membership could be updated, or other actions may be taken as necessary.
For further explanation, FIG. 7 sets forth diagrams of metadata representations that may be implemented as a structured collection of metadata objects that, together, may represent a logical volume of storage data, or a portion of a logical volume, in accordance with some embodiments of the present disclosure. Metadata representations 750, 754, and 760 may be stored within a storage system (706), and one or more metadata representations may be generated and maintained for each of multiple storage objects, such as volumes, or portions of volumes, stored within a storage system (706).
While other types of structured collections of the metadata objects are possible, in this example, metadata representations may be structured as a directed acyclic graph (DAG) of nodes, where, to maintain efficient access to any given node, the DAG may be structured and balanced according to various methods. For example, a DAG for a metadata representation may be defined as a type of B-tree, and balanced accordingly in response to changes to the structure of the metadata representation, where changes to the metadata representation may occur in response to changes to, or additions to, underlying data represented by the metadata representation. While in this example, there are only two levels for the sake of simplicity, in other examples, metadata representations may span across multiple levels and may include hundreds or thousands of nodes, where each node may include any number of links to other nodes.
Further, in this example, the leaves of a metadata representation may include pointers to the stored data for a volume, or portion of a volume, where a logical address, or a volume and offset, may be used to identify and navigate through the metadata representation to reach one or more leaf nodes that reference stored data corresponding to the logical address. For example, a volume (752) may be represented by a metadata representation (750), which includes multiple metadata object nodes (752, 752A-752N), where leaf nodes (752A-752N) include pointers to respective data objects (753A-753N, 757). Data objects may be any size unit of data within a storage system (706). For example, data objects (753A-753N, 757) may each be a logical extent, where logical extents may be some specified size, such as 1 MB, 4 MB, or some other size.
In this example, a snapshot (756) may be created as a snapshot of a storage object, in this case, a volume (752), where at the point in time when the snapshot (756) is created, the metadata representation (754) for the snapshot (756) includes all of the metadata objects for the metadata representation (750) for the volume (752). Further, in response to creation of the snapshot (756), the metadata representation (754) may be designated to be read only. However, the volume (752) sharing the metadata representation may continue to be modified, and while at the moment the snapshot is created, the metadata representations for the volume (752) and the snapshot (756) are identical, as modifications are made to data corresponding to the volume (752), and in response to the modifications, the metadata representations for the volume (752) and the snapshot (756) may diverge and become different.
For example, given a metadata representation (750) to represent a volume (752) and a metadata representation (754) to represent a snapshot (756), the storage system (706) may receive an I/O operation that writes to data that is ultimately stored within a particular data object (753B), where the data object (753B) is pointed to by a leaf node pointer (752B), and where the leaf node pointer (752B) is part of both metadata representations (750, 754). In response to the write operation, the read only data objects (753A-753N) referred to by the metadata representation (754) remain unchanged, and the pointer (752B) may also remain unchanged. However, the metadata representation (750), which represents the current volume (752), is modified to include a new data object to hold the data written by the write operation, where the modified metadata representation is depicted as the metadata representation (760). Further, the write operation may be directed to only a portion of the data object (753B), and consequently, the new data object (757) may include a copy of previous contents of the data object (753B) in addition to the payload for the write operation.
In this example, as part of processing the write operation, the metadata representation (760) for the volume (752) is modified to remove an existing metadata object pointer (752B) and to include a new metadata object pointer (758), where the new metadata object pointer (758) is configured to point to a new data object (757), where the new data object (757) stores the data written by the write operation. Further, the metadata representation (760) for the volume (752) continues to include all metadata objects included within the previous metadata representation (750)—with the exclusion of the metadata object pointer (752B) that referenced the target data object, where the metadata object pointer (752B) continues to reference the read only data object (753B) that would have been overwritten.
In this way, using metadata representations, a volume or a portion of a volume may be considered to be snapshotted, or considered to be copied, by creating metadata objects, and without actual duplication of data objects—where the duplication of data objects may be deferred until a write operation is directed at one of the read only data objects referred to by the metadata representations.
In other words, an advantage of using a metadata representation to represent a volume is that a snapshot or a copy of a volume may be created and be accessible in constant order time, and specifically, in the time it takes to create a metadata object for the snapshot or copy, and to create a reference for the snapshot or copy metadata object to the existing metadata representation for the volume being snapshotted or copied.
As an example use, a virtualized copy-by-reference may make use of a metadata representation in a manner that is similar to the use of a metadata representation in creating a snapshot of a volume—where a metadata representation for a virtualized copy-by-reference may often correspond to a portion of a metadata representation for an entire volume. An example implementation of virtualized copy-by-reference may be within the context of a virtualized storage system, where multiple block ranges within and between volumes may reference a unified copy of stored data. In such virtualized storage system, the metadata described above may be used to handle the relationship between virtual, or logical, addresses and physical, or real, addresses—in other words, the metadata representation of stored data enables a virtualized storage system that may be considered flash-friendly in that it reduces, or minimizes, wear on flash memory.
In some examples, logical extents may be combined in various ways, including as simple collections or as logically related address ranges within some larger-scale logical extent that is formed as a set of logical extent references. These larger combinations could also be given logical extent identities of various kinds, and could be further combined into still larger logical extents or collections. A copy-on-write status could apply to various layers, and in various ways depending on the implementation. For example, a copy on write status applied to a logical collection of logical collections of extents might result in a copied collection retaining references to unchanged logical extents and the creation of copied-on-write logical extents (through copying references to any unchanged stored data blocks as needed) when only part of the copy-on-write logical collection is changed.
Deduplication, volume snapshots, or block range snapshots may be implemented in this model through combinations of referencing stored data blocks, or referencing logical extents, or marking logical extents (or identified collections of logical extents) as copy-on-write.
Further, with flash storage systems, stored data blocks may be organized and grouped together in various ways as collections are written out into pages that are part of larger erase blocks. Eventual garbage collection of deleted or replaced stored data blocks may involve moving content stored in some number of pages elsewhere so that an entire erase block can be erased and prepared for reuse. This process of selecting physical flash pages, eventually migrating and garbage collecting them, and then erasing flash erase blocks for reuse may or may not be coordinated, driven by, or performed by the aspect of a storage system that is also handling logical extents, deduplication, compression, snapshots, virtual copying, or other storage system functions. A coordinated or driven process for selecting pages, migrating pages, garbage collecting and erasing erase blocks may further take into account various characteristics of the flash memory device cells, pages, and erase blocks such as number of uses, aging predictions, adjustments to voltage levels or numbers of retries needed in the past to recover stored data. They may also take into account analysis and predictions across all flash memory devices within the storage system.
To continue with this example, where a storage system may be implemented based on directed acyclic graphs comprising logical extents, logical extents can be categorized into two types: leaf logical extents, which reference some amount of stored data in some way, and composite logical extents, which reference other leaf or composite logical extents.
A leaf extent can reference data in a variety of ways. It can point directly to a single range of stored data (e.g., 64 kilobytes of data), or it can be a collection of references to stored data (e.g., a 1 megabyte “range” of content that maps some number of virtual blocks associated with the range to physically stored blocks). In the latter case, these blocks may be referenced using some identity, and some blocks within the range of the extent may not be mapped to anything. Also, in that latter case, these block references need not be unique, allowing multiple mappings from virtual blocks within some number of logical extents within and across some number of volumes to map to the same physically stored blocks. Instead of stored block references, a logical extent could encode simple patterns: for example, a block which is a string of identical bytes could simply encode that the block is a repeated pattern of identical bytes.
A composite logical extent can be a logical range of content with some virtual size, which comprises a plurality of maps that each map from a subrange of the composite logical extent logical range of content to an underlying leaf or composite logical extent. Transforming a request related to content for a composite logical extent, then, involves taking the content range for the request within the context of the composite logical extent, determining which underlying leaf or composite logical extents that request maps to, and transforming the request to apply to an appropriate range of content within those underlying leaf or composite logical extents.
Volumes, or files or other types of storage objects, can be described as composite logical extents. Thus, these presented storage objects can be organized using this extent model.
Depending on implementation, leaf or composite logical extents could be referenced from a plurality of other composite logical extents, effectively allowing inexpensive duplication of larger collections of content within and across volumes. Thus, logical extents can be arranged essentially within an acyclic graph of references, each ending in leaf logical extents. This can be used to make copies of volumes, to make snapshots of volumes, or as part of supporting virtual range copies within and between volumes as part of EXTENDED COPY or similar types of operations.
An implementation may provide each logical extent with an identity which can be used to name it. This simplifies referencing, since the references within composite logical extents become lists comprising logical extent identities and a logical subrange corresponding to each such logical extent identity. Within logical extents, each stored data block reference may also be based on some identity used to name it.
To support these duplicated uses of extents, we can add a further capability: copy-on-write logical extents. When a modifying operation affects a copy-on-write leaf or composite logical extent the logical extent is copied, with the copy being a new reference and possibly having a new identity (depending on implementation). The copy retains all references or identities related to underlying leaf or composite logical extents, but with whatever modifications result from the modifying operation. For example, a WRITE, WRITE SAME, XDWRITEREAD, XPWRITE, or COMPARE AND WRITE request may store new blocks in the storage system (or use deduplication techniques to identify existing stored blocks), resulting in modifying the corresponding leaf logical extents to reference or store identities to a new set of blocks, possibly replacing references and stored identities for a previous set of blocks. Alternately, an UNMAP request may modify a leaf logical extent to remove one or more block references. In both types of cases, a leaf logical extent is modified. If the leaf logical extent is copy-on-write, then a new leaf logical extent will be created that is formed by copying unaffected block references from the old extent and then replacing or removing block references based on the modifying operation.
A composite logical extent that was used to locate the leaf logical extent may then be modified to store the new leaf logical extent reference or identity associated with the copied and modified leaf logical extent as a replacement for the previous leaf logical extent. If that composite logical extent is copy-on-write, then a new composite logical extent is created as a new reference or with a new identity, and any unaffected references or identities to its underlying logical extents are copied to that new composite logical extent, with the previous leaf logical extent reference or identity being replaced with the new leaf logical extent reference or identity.
This process continues further backward from referenced extent to referencing composite extent, based on the search path through the acyclic graph used to process the modifying operation, with all copy-on-write logical extents being copied, modified, and replaced.
These copied leaf and composite logical extents can then drop the characteristic of being copy on write, so that further modifications do not result in an additional copy. For example, the first time some underlying logical extent within a copy-on-write “parent” composite extent is modified, that underlying logical extent may be copied and modified, with the copy having a new identity which is then written into a copied and replaced instance of the parent composite logical extent. However, a second time some other underlying logical extent is copied and modified and with that other underlying logical extent copy's new identity being written to the parent composite logical extent, the parent can then be modified in place with no further copy and replace necessary on behalf of references to the parent composite logical extent.
Modifying operations to new regions of a volume or of a composite logical extent for which there is no current leaf logical extent may create a new leaf logical extent to store the results of those modifications. If that new logical extent is to be referenced from an existing copy-on-write composite logical extent, then that existing copy-on-write composite logical extent will be modified to reference the new logical extent, resulting in another copy, modify, and replace sequence of operations similar to the sequence for modifying an existing leaf logical extent.
If a parent composite logical extent cannot be grown large enough (based on implementation) to cover an address range associated that includes new leaf logical extents to create for a new modifying operation, then the parent composite logical extent may be copied into two or more new composite logical extents which are then referenced from a single “grandparent” composite logical extent which yet again is a new reference or a new identity. If that grandparent logical extent is itself found through another composite logical extent that is copy-on-write, then that another composite logical extent will be copied and modified and replaced in a similar way as described in previous paragraphs. This copy-on-write model can be used as part of implementing snapshots, volume copies, and virtual volume address range copies within a storage system implementation based on these directed acyclic graphs of logical extents. To make a snapshot as a read-only copy of an otherwise writable volume, a graph of logical extents associated with the volume is marked copy-on-write and a reference to the original composite logical extents are retained by the snapshot. Modifying operations to the volume will then make logical extent copies as needed, resulting in the volume storing the results of those modifying operations and the snapshots retaining the original content. Volume copies are similar, except that both the original volume and the copied volume can modify content resulting in their own copied logical extent graphs and subgraphs.
Virtual volume address range copies can operate either by copying block references within and between leaf logical extents (which does not itself involve using copy-on-write techniques unless changes to block references modifies copy-on-write leaf logical extents). Alternately, virtual volume address range copies can duplicate references to leaf or composite logical extents, which works well for volume address range copies of larger address ranges. Further, this allows graphs to become directed acyclic graphs of references rather than merely reference trees. Copy-on-write techniques associated with duplicated logical extent references can be used to ensure that modifying operations to the source or target of a virtual address range copy will result in the creation of new logical extents to store those modifications without affecting the target or the source that share the same logical extent immediately after the volume address range copy operation.
Input/output operations for pods may also be implemented based on replicating directed acyclic graphs of logical extents. For example, each storage system within a pod could implement private graphs of logical extents, such that the graphs on one storage system for a pod have no particular relationship to the graphs on any second storage system for the pod. However, there is value in synchronizing the graphs between storage systems in a pod. This can be useful for resynchronization and for coordinating features such as asynchronous or snapshot based replication to remote storage systems. Further, it may be useful for reducing some overhead for handling the distribution of snapshot and copy related processing. In such a model, keeping the content of a pod in sync across all in-sync storage systems for a pod is essentially the same as keeping graphs of leaf and composite logical extents in sync for all volumes across all in-sync storage systems for the pod, and ensuring that the content of all logical extents is in-sync. To be in sync, matching leaf and composite logical extents should either have the same identity or should have mappable identities. Mapping could involve some set of intermediate mapping tables or could involve some other type of identity translation. In some cases, identities of blocks mapped by leaf logical extents could also be kept in sync.
In a pod implementation based on a leader and followers, with a single leader for each pod, the leader can be in charge of determining any changes to the logical extent graphs. If a new leaf or composite logical extent is to be created, it can be given an identity. If an existing leaf or composite logical extent is to be copied to form a new logical extent with modifications, the new logical extent can be described as a copy of a previous logical extent with some set of modifications. If an existing logical extent is to be split, the split can be described along with the new resulting identities. If a logical extent is to be referenced as an underlying logical extent from some additional composite logical extent, that reference can be described as a change to the composite logical extent to reference that underlying logical extent.
Modifying operations in a pod thus comprises distributing descriptions of modifications to logical extent graphs (where new logical extents are created to extend content or where logical extents are copied, modified, and replaced to handle copy-on-write states related to snapshots, volume copies, and volume address range copies) and distributing descriptions and content for modifications to the content of leaf logical extents. An additional benefit that comes from using metadata in the form of directed acyclic graphs, as described above, is that I/O operations that modify stored data in physical storage may be given effect at a user level through the modification of metadata corresponding to the stored data in physical storage—without modifying the stored data in physical storage. In the disclosed embodiments of storage systems, where the physical storage may be a solid state drive, the wear that accompanies modifications to flash memory may be avoided or reduced due to I/O operations being given effect through the modifications of the metadata representing the data targeted by the I/O operations instead of through the reading, erasing, or writing of flash memory. Further, as noted above, in such a virtualized storage system, the metadata described above may be used to handle the relationship between virtual, or logical, addresses and physical, or real, addresses—in other words, the metadata representation of stored data enables a virtualized storage system that may be considered flash-friendly in that it reduces, or minimizes, wear on flash memory.
Leader storage systems may perform their own local operations to implement these descriptions in the context of their local copy of the pod dataset and the local storage system's metadata. Further, the in-sync followers perform their own separate local operations to implement these descriptions in the context of their separate local copy of the pod dataset and their separate local storage system's metadata. When both leader and follower operations are complete, the result is compatible graphs of logical extents with compatible leaf logical extent content. These graphs of logical extents then become a type of “common metadata” as described in previous examples. This common metadata can be described as dependencies between modifying operations and required common metadata. Transformations to graphs can be described as separate operations within a set of or more predicates that may describe relationships, such as dependencies, with one or more other operations. In other words, interdependencies between operations may be described as a set of precursors that one operation depends on in some way, where the set of precursors may be considered predicates that must be true for an operation to complete. A fuller description of predicates may be found within application Reference Ser. No. 15/696,418, which is included herein by reference in its entirety. Alternately, each modifying operation that relies on a particular same graph transformation that has not yet been known to complete across the pod can include the parts of any graph transformation that it relies on. Processing an operation description that identifies a “new” leaf or composite logical extent that already exists can avoid creating the new logical extent since that part was already handled in the processing of some earlier operation, and can instead implement only the parts of the operation processing that change the content of leaf or composite logical extents. It is a role of the leader to ensure that transformations are compatible with each other. For example, we can start with two writes come that come in for a pod. A first write replaces a composite logical extent A with a copy of formed as composite logical extent B, replaces a leaf logical extent C with a copy as leaf logical extent D and with modifications to store the content for the second write, and further writes leaf logical extent D into composite logical extent B. Meanwhile, a second write implies the same copy and replacement of composite logical extent A with composite logical extent B but copies and replaces a different leaf logical extent E with a logical extent F which is modified to store the content of the second write, and further writes logical extent F into logical extent B. In that case, the description for the first write can include the replacement of A with B and C with D and the writing of D into composite logical extent B and the writing of the content of the first write into leaf extend B; and, the description of the second write can include the replacement of A with B and E with F and the writing of F into composite logical extent B, along with the content of the second write which will be written to leaf extent F. A leader or any follower can then separately process the first write or the second write in any order, and the end result is B copying and replacing A, D copying and replacing C, F copying replacing E, and D and F being written into composite logical extent B. A second copy of A to form B can be avoided by recognizing that B already exists. In this way, a leader can ensure that the pod maintains compatible common metadata for a logical extent graph across in-sync storage systems for a pod.
Given an implementation of storage systems using directed acyclic graphs of logical extents, recovery of pods based on replicated directed acyclic graphs of logical extents may be implemented. Specifically, in this example, recovery in pods may be based on replicated extent graphs then involves recovering consistency of these graphs as well as recovering content of leaf logical extents. In this implementation of recovery, operations may include querying for graph transformations that are not known to have completed on all in-sync storage systems for a pod, as well as all leaf logical extent content modifications that are not known to have completed across all storage systems for the pod. Such querying could be based on operations since some coordinated checkpoint, or could simply be operations not known to have completed where each storage system keeps a list of operations during normal operation that have not yet been signaled as completed. In this example, graph transformations are straightforward: a graph transformation may create new things, copy old things to new things, and copy old things into two or more split new things, or they modify composite extents to modify their references to other extents. Any stored operation description found on any in-sync storage system that creates or replaces any logical extent can be copied and performed on any other storage system that does not yet have that logical extent. Operations that describe modifications to leaf or composite logical extents can apply those modifications to any in-sync storage system that had not yet applied them, as long as the involved leaf or composite logical extents have been recovered properly.
In another example, as an alternative to using a logical extent graph, storage may be implemented based on a replicated content-addressable store. In a content-addressable store, for each block of data (for example, every 512 bytes, 4096 bytes, 8192 bytes or even 16384 bytes) a unique hash value (sometimes also called a fingerprint) is calculated, based on the block content, so that a volume or an extent range of a volume can be described as a list of references to blocks that have a particular hash value. In a synchronously replicated storage system implementation based on references to blocks with the same hash value, replication could involve a first storage system receiving blocks, calculating fingerprints for those blocks, identifying block references for those fingerprints, and delivering changes to one or a plurality of additional storage systems as updates to the mapping of volume blocks to referenced blocks. If a block is found to have already been stored by the first storage system, that storage system can use its reference to name the reference in each of the additional storage systems (either because the reference uses the same hash value or because an identifier for the reference is either identical or can be mapped readily). Alternately, if a block is not found by the first storage system, then content of the first storage system may be delivered to other storage systems as part of the operation description along with the hash value or identity associated with that block content. Further, each in-sync storage system's volume descriptions are then updated with the new block references. Recovery in such a store may then include comparing recently updated block references for a volume. If block references differ between different in-sync storage systems for a pod, then one version of each reference can be copied to other storage systems to make them consistent. If the block reference on one system does not exist, then it be copied from some storage system that does store a block for that reference. Virtual copy operations can be supported in such a block or hash reference store by copying the references as part of implementing the virtual copy operation.
For further explanation, FIG. 4 sets forth a flow chart illustrating an example method for synchronizing metadata among storage systems synchronously replicating a dataset according to some embodiments of the present disclosure. Although depicted in less detail, the storage systems (402, 404, 406) depicted in FIG. 4 may be similar to the storage systems described above with reference to FIGS. 1A-1D, FIGS. 2A-2G, FIGS. 3A-3C, or any combination thereof. In fact, the storage systems (402, 404, 406) depicted in FIG. 4 may include the same, fewer, additional components as the storage systems described above.
As described above, metadata may be synchronized among storage systems that are synchronously replicating a dataset. Such metadata may be referred to as common metadata, or shared metadata, that is stored by a storage system on behalf of a pod related to the mapping of segments of content stored within the pod to virtual address within storage objects within the pod, where information related to those mappings is synchronized between member storage systems for the pod to ensure correct behavior—or better performance—for storage operations related to the pod. In some examples, a storage object may implement a volume or a snapshot. The synchronized metadata may include: (a) information to keep volume content mappings synchronized among the storage systems in the pod; (b) tracking data for recovery checkpoints or for in-progress write operations; (c) information related to the delivery of data and mapping information to a remote storage system for asynchronous or periodic replication.
Information to keep volume content mappings synchronized among the storage systems in the pod may enable efficient creating of snapshots, which in turn enables that subsequent updates, copies of snapshots, or snapshot removals may be performed efficiently and consistently across the pod member storage systems.
Tracking data for recovery checkpoints or for in-progress write operations may enable efficient crash recovery and efficient detection of content or volume mappings that may have been partially or completely applied on individual storage systems for a pod, but that may not have been completely applied on other storage systems for the pod.
Information related to the delivery of data and mapping information to a remote storage system for asynchronous or periodic replication may enable more than one member storage system for a pod to serve as a source for the replicated pod content with minimal concerns for dealing with mismatches in mapping and differencing metadata used to drive asynchronous or periodic replication.
In some examples, shared metadata may include descriptions for, or indications of, a named grouping, or identifiers for, of one or more volumes or one or more storage objects that are a subset of an entire synchronously replicated dataset for a pod—where such a of volumes or storage objects of a dataset may be referred to as a consistency group. A consistency group may be defined to specify a subset of volumes or storage objects of the dataset to be used for consistent snapshots, asynchronous replication, or periodic replication. In some examples, a consistency group may be calculated dynamically, such as by including all volumes connected to a particular set of hosts or host network ports, or that are connected to a particular set of applications or virtual machines or containers, where the applications, virtual machines, or containers may operate on external server systems or may operate on one or more of the storage systems that are members of a pod. In other examples, a consistency group may be defined according to user selections of a type of data or set of data, or specifications of a consistency group similar to the dynamic calculation, where a user may specify, for example through a command or management console, that a particular, or named, consistency group be created to include all volumes connected to a particular set of hosts or host network ports, or be created to include data for a particular set of applications or virtual machines or containers.
In an example using a consistency group, a first consistency group snapshot of a consistency group may include a first set of snapshots for all volumes or other storage objects that are members of the consistency group at the time of the first dataset snapshot, with a second consistency group snapshot of the same consistency group including a second set of snapshots for the volumes or other storage objects that are members of the consistency group at the time of the second dataset snapshot. In other examples, a snapshot of the dataset may be stored on one or more target storage systems in an asynchronous manner. Similarly, asynchronous replication of a consistency group may account for dynamic changes to member volumes and other storage objects of the consistency group, where consistency group snapshots of the consistency group at either the source or the target of the asynchronous replication link include the volumes and other storage objects that are members in relationship to the consistency group at the time that the dataset snapshot relates to. In the case of a target of an asynchronous replication connection, the time that the dataset snapshot relates to depends on the dynamic dataset of the sender as it was received and was in process at the time of the consistency group snapshot on the target. For example, if a target of an asynchronous replication is, say, 2000 operations behind, where some of those operations are consistency group member changes, where a first set of such changes are more than 2000 operations ago for the source, and a second set of changes are within the last 2000, then a consistency group snapshot at that time on the target will account for the first set of member changes and will not account for the second set of changes. Other uses of the target of asynchronous replication may similarly account for the nature of the time of the dataset for the consistency group in determining the volumes or other storage objects (and their content) for those uses. For example, in the same case of asynchronous replication being 2000 operations behind, use of the target for a disaster recovery failover might start from a dataset that includes the volumes and other storage objects (and their content) as they were 2000 operations ago at the source. In this discussion, concurrent operations at the source (e.g., writes, storage object creations or deletions, changes to properties that affect inclusion or exclusion of volumes or other storage objects or other data from a consistency group, or other operations that were in progress and not signaled as completed at a same point in time) might not have a single well-defined ordering, so the count of operations only needs to represent some plausible ordering based on any allowed ordering of concurrent operations on the source.
As another example using consistency groups, in the case of periodic replication based on replication of consistency group snapshots, each replicated consistency group snapshot would include the volumes and other storage objects at the time each consistency group snapshot was formed on the source. Ensuring that membership in a consistency group is kept consistent by using common, or shared, metadata, ensures that a fault—or other change which may cause the source of replication, or the system that forms a dataset snapshot, to switch from one storage system in a pod to another—does not lose information needed for properly handling those consistency group snapshots or the consistency group replication. Further, this type of handling may allow for multiple storage systems that are members of a pod to concurrently serve as source systems for asynchronous or periodic replication.
Further, synchronized metadata describing mapping of segments to storage objects is not limited to mappings themselves, and may include additional information such as sequence numbers (or some other value for identifying stored data), timestamps, volume/snapshot relationships, checkpoint identities, trees or graphs defining hierarchies, or directed graphs of mapping relationships, among other storage system information.
For further explanation, FIG. 8 sets forth a diagram of a computing environment supporting an implementation of asynchronous replication of synchronously replicated data in accordance with some embodiments of the present disclosure.
As discussed above with regard to FIGS. 4-7, a synchronously replicated dataset is resilient to multiple types of failures in part due to being replicated among more than a single storage system. However, data resiliency may be further improved by replicating a dataset that is being synchronously replicated among storage systems to a storage system that is not part of the storage systems that are synchronously replicating the dataset. In other words, using the terminology introduced above, a dataset within a pod may be asynchronously replicated to one or more storage systems that are not members of the pod. However, in some examples, because a pod is a logical construct that is independent from the physical storage devices implementing the pod, a synchronously replicated dataset within the pod may be asynchronously replicated to one or more storage devices that also implement the pod.
In the example computing environment depicted in FIG. 8, multiple storage systems (800A-800N) may synchronously replicate a dataset (804) within a pod (802), where another storage system (820) is connected over a network (801) to the multiple storage systems (800A-800N), but where the other storage system (820) is not a member of the multiple storage systems (800A-800N) that are synchronously replicating the dataset (804). In this example, the multiple storage systems (800A-800N) may be referred to as source storage systems and the other storage system (820) may be referred to as a target storage system.
Further, in this example, because the dataset (804) being replicated is synchronously replicated among the multiple storage systems (800A-800N), portions of the dataset (804) may be asynchronously copied from more than one of the storage systems (800A-800N) either serially, in parallel, or both serially and in parallel. As depicted in FIG. 8, a given storage system among the multiple storage systems may establish one or more connections to a target storage system (820), where each connection may transmit a respective portion of the dataset (804). In this example, a first storage system (800A) may include source connections (806A-806P), and an Nth storage system (800N) of the active cluster of storage systems (800A-800N) may include source connections (810A-810Q)—where source connections (806A-806P) transmit respective portions of dataset (804), depicted as subsets (808A-808P), and so on for each storage system until the Nth storage system (800N), where source connections (810A-810Q) transmit respective portions of dataset (804), depicted as subsets (812A-812Q).
Correspondingly, in this example, the target storage system (820) may include a respective target connection for each of the connections on the multiple storage systems, where the target connections (820A-820P, . . . , 824A-824Q) may respectively receive a portion of the dataset (804), where the respective portions are depicted as subsets (808A-808P, . . . , 812A-812Q). In this example, a given connection may be a network layer connection, such as a TCP connection; however, in other examples, and in general, any type of link or connection that may reliably transport data across a network may be used.
Further, the extent to which each storage system (800A-800N) is responsible for replicating a portion of the dataset (804) may be based on a variety of factors, including one or more of: physical proximity, where closer storage systems may be assigned a greater portion of the dataset (804) to replicate as compared to other source storage systems; source storage system performance characteristics, where storage systems with greater computational or physical resources may be assigned a greater portion of the dataset (804) to replicate as compared to other source storage systems; source storage systems workloads or busyness, including expected workloads or busyness, where source storage systems with greater workload or busyness or greater expected workloads or busyness may be assigned a smaller portion of the dataset (804) to replicate as compared to other source storage systems; network bandwidth availability, where source storage systems with greater available bandwidth availability are assigned a greater portion of the dataset (804) to replicate as compared to other source storage systems; network latency characteristics, where source storage systems with smaller network latencies are assigned a greater portion of the dataset (804) to replicate as compared to other source storage systems; including other factors that may affect storage system performance in replicating one or more portions of the dataset (804).
In other examples, a rule that specifies which portions of a dataset are replicated by which storage systems may be defined by a user. Generally, multiple rules may be triggered under different circumstances, or under specific combinations of events. For example, a rule may specify that one or more source storage systems have their network bandwidth limited, or throttled, in accordance with a rule specifying a particular balance of network loads among the set of source storage systems. In some examples, a rule may specify that a given replication process for a given dataset not consume more than X % of network bandwidth on given storage systems, where each storage systems may be limited according to a respective percentage or quantity of network data over a period of time. In other examples, a controller on the source storage system may implement one or more rules for one or more other storage systems based on telemetry data, or performance metrics, for one or more of the source storage systems, where the telemetry data, or performance metrics, indicate an expected future load or event that may consume computing resources. For example, a given storage system may have a scheduled workload activity, and to reduce the impact of the schedule workload activity on a replication process, and in dependence upon the scheduled workload activity or telemetry data, a controller may reduce the replication load on the given storage system and increase the replication load on one or more of the remaining source storage systems.
In some examples, replication of the dataset (804) may continue in the event of one or more storage systems (800A-800N) falling out of synchronization. For example, if the dataset (804) is a portion of the synchronized data being storage among the multiple storage systems (800A-800N), and the dataset (804) is synchronized while other data may not be synchronized, then the replication of the dataset (804) may continue without the dataset (804) becoming inconsistent. In the case of the dataset (804) being a snapshot, if two storage systems are in-sync when a snapshot is taken, and then the two storage systems fall out-of-sync, if the target storage system is pulling from the most recent snapshot on the two storage systems, then the target storage system may continue replicating the snapshot if the most recent snapshot is in-sync.
With reference to FIG. 8, in the example of the dataset (804) being a snapshot, a member of the active cluster, such as storage system (800A) among storage systems (800A-800N), may notify the target storage system (820) that a snapshot has been taken, where storage system (800A) may send the notification based on being a leader. A leader is described in greater detail above with reference to FIGS. 4-7, however, in other examples, the active cluster, storage systems (800A-800N), may have no particular leader, and one or more of the storage systems (800A-800N) may send a notification to the target storage system (820) in response to the dataset being ready for replication, such as a snapshot having been successfully taken.
As depicted in FIG. 8, a dataset (804) may be split into multiple subsets (808A-808P, . . . , 812A-812Q) transmitted from multiple source connections (806A-806P, . . . , 810A-810Q) to multiple target connections (820A-820P, . . . , 824A-824Q) via multiple target connections (820A-820P, . . . , 824A-824Q) over a network (801), where as discussed below, different mappings between source connections and target sessions are possible.
Continuing with this example, in response to the target storage system (820) receiving a notification that a dataset is ready to be replicated from one or more of the storage systems (800A-800N), the target storage system (820) may schedule multiple replication sessions to fetch metadata and/or data corresponding to the dataset (804) from one or more of the source storage systems (800A-800N) providing a portion, or subset, of the dataset (804). In this example, each session may correspond to a previously established connection, such as a long-lived TCP connection, where in some examples, multiple sessions may be multiplexed onto each connection.
Further, in this example, the target storage system (820) may create a work queue that includes a list of each portion, or subset, of data that is to be fetched, or received, for the dataset (804). While in this example, a queue is implemented, in other examples, other data structures may be used or created, where the data structure may include dataset information regarding the dataset (804) that is created in response to receiving a notification of the dataset (804) to be transferred, where the dataset information may describe characteristics of the dataset, such as address space(s), size, or other characteristics of the dataset (804). In this example, a given session may pull, or request, data in accordance with a portion, or subset, of data described within a work queue entry.
In some examples, a given work queue entry may also identify a source storage system (800A-800N) from which to pull, or request, data, where the source storage system may be determined when the queue entry is made, and where the determination of the source storage system may be based on telemetry, or performance metrics or characteristics, of the source storage systems (800A-800N). In other examples, a work queue entry may simply identify a subset of the dataset (804), where the source storage system (800A-800N) is determined dynamically when the work queue entry is retrieved, and where the dynamic determination of the source storage system may be made based on telemetry data for one or more of the source storage system, or performance metrics or characteristics, of the source storage systems (800A-800N), or based on current network conditions, or a combination of these factors.
Continuing with this example, a given session among the multiple sessions, having accessed a work queue entry, may then handle requesting, pulling, or fetching data corresponding to the work queue entry and, in response to receiving the requested data, write the data within the target storage system (840). In some examples, each session may work on several items of work corresponding to several different work queue entries, where each session may work on more than one item of work simultaneously. In some examples, a session may form a pipeline with several stages, such as looking up metadata on the source, checking for duplicates on the target, and fetching data from the source—where tasks corresponding to each stage of the pipeline may be performed independently of each other.
However, in some examples, the number of work items that a given session may handle may be limited, such as by the number of stages in a pipeline. In some examples, responsive to a session claiming a maximum amount of work items on which the session may work simultaneously, the session may be prevented from claiming additional work items from the work queue—a constraint that when applied to all sessions provides a flow control mechanism to the multiple sessions. In this way, work items in the work queue may self-balance among the multiple sessions, which results in balancing network traffic among the one or more storage systems replicating the dataset (804)—where in some cases, the multiple sessions may be using different TCP connections, including different physical network interfaces and/or different network paths. For example, if a particular session is slow, the flow control mechanism serves to limit the number of work items the particular session may obtain from the work queue. Similarly, faster sessions may claim and process a quantity of work items from the work queue at a speed at which the faster sessions are operating, thereby preventing the faster sessions from becoming idle. Another benefit of the flow control mechanism described is that if any given session stalls or fails, a monitoring process on the target storage system may return the work item for the stalled or failed session into the work queue, where a stall or fail may be determined by a timeout event for the given session. In some embodiments, an alternative way of considering balancing work items among the multiple sessions is the use of a “shortest queue” priority model, where work is given to whichever network path has the least currently outstanding work. Continuing with this example, if paths complete work at the same rate, then this results in an even load across all paths; however, if some paths complete work faster—for whatever reason, including less load on a given controller, a shorter network link, a better network link, temporary network congestion, among other reasons—then more work is given to paths that complete work faster.
In some examples, the work queue may be populated by one or more of the sessions. Further, in some examples, a work item may be a small chunk of data or metadata, such as a 1 MB chunk, or the work item may be a larger chunk of logical space. In some examples, each work item may correspond to a node or extent within a BDAG representation of a dataset, or snapshot, such as the BDAG representation described above with reference to FIG. 7. In other examples, in dependence upon a size of a chunk, portion, or subset, of data to transfer, the session may either request or fetch and write the complete quantity of data or make multiple requests or fetches to retrieve the data in smaller parts. For example, if the size of a chunk is small, a single request or fetch may be made to retrieve the data for the chunk, however, if the size of the chunk is large, then multiple requests or fetches may be made to retrieve the data for the chunk—where small or large may be determined according to a threshold value, such as a 1 MB, 16 MB, or some other specified number of bytes.
In some implementations, a cost for handling different work items from the work queue may be different, where cost may be measured in terms of expected time to complete, storage space, network bandwidth, processing cycles, or some other resource metric such as time to fetch an extent or break down a node into smaller nodes. In this example, a given session may estimate a cost of completing a work item, and as work proceeds through the given session's pipeline, the estimate may be updated, which results in allowing a session to accept additional work items before completing a current work item.
For further explanation, FIG. 9 sets forth a flow chart illustrating steps that may be performed to asynchronously replicate a synchronously replicated dataset according to some embodiments of the present disclosure. Although depicted in less detail, the storage systems (800A-800N, 820) depicted in FIG. 9 may be similar to the storage systems described above with reference to FIGS. 1A-1D, FIGS. 2A-2G, FIGS. 3A-3B, and FIGS. 4-7, or any combination thereof. In fact, the storage systems (800A-800N, 820) depicted in FIG. 9 may include the same, fewer, additional components as the storage systems described above.
In the example method depicted in FIG. 9, a target storage system (820) may determine (902), at the target storage system (820), multiple work items corresponding to a dataset (804) stored among multiple source storage systems (800A-800N), where each respective work item corresponds to a respective subset of the dataset (804). Determining (902), at the target storage system (820), the multiple work items corresponding to the dataset (804) stored among the multiple source storage systems (800A-800N) may be implemented by a controller of the target storage system (820) receiving a notification from one or more of the source storage systems (800A-800N) indicating that a dataset is ready for replication. In this example, in response to receiving the notification, the target storage system (820) may request information describing the dataset (804) that may be used to identify and transfer the dataset (804), where the information describing the dataset (804) may be a metadata structure as described above with reference to FIGS. 7 and 8. For example, as described above with reference to FIG. 8, the metadata structure may include nodes that correspond to blocks of the dataset, and where the target storage system (820) may create work items for each block to be transferred; however, in other examples, a work item may correspond to other divisions or subsets of the dataset (804).
The example method depicted in FIG. 9 also includes, for (904) each session from among a plurality of sessions (822A-822P, 826A-826Q) operating on the target storage system (820): determining (906) one or more computing environment factors (953) affecting performance of the replication of data from one or more of the multiple source storage systems (800A-800N) to the target storage system (820); identifying (908), for a given session and based on the one or more computing environment factors (953), a respective source storage system and a quantity of work items; and replicating (910), from the respective source storage system, one or more subsets of data corresponding to the quantity of work items.
Determining (906) the one or more computing environment factors (953) affecting performance of the replication of data from on or more of the multiple source storage systems (800A-800N) to the target storage system (820) may be implemented by the target storage system (820) receiving one or more of: performance metrics for one or more of the source storage systems (800A-800N), performance metrics for one or more other storage systems that are configured similarly to the source storage systems (800A-800N), telemetry data indicating current or expected workloads, or metrics describing current or expected network traffic conditions between one or more of the source storage systems (800A-800N) and the target storage system (820). In some examples, the computing environment factors (953) may be received after replication of a dataset (804) has already begun, and in such a case, the target storage system (820) may dynamically adjust, or rebalance, replication workloads for each of the source storage systems in dependence upon the computing environment factors (953) received or determined after replication has begun. In other examples, computing environment factors (953) may include quality-of-service (QOS) metrics for workloads that are operating on a given storage system, where the throughput may be limited or modulated in accordance with QOS metrics, including allowing replication when more cycles are available from a workload that is below a QOS metric or guarantee, while replication is fully able to make use of cycles that are available beyond the QOS metric or guarantee for the storage system QOS-affected workloads.
Identifying (908), for a given session and based on or more of the computing environment factors (953), a respective source storage system and a quantity of work items (956) may be implemented as described above with reference to FIG. 8, where a session may remove one or more work items (956) from a work queue (830) for replicating one or more subsets of the dataset (804), for example, in dependence upon network performance for a corresponding connection to a corresponding storage system, or in dependence upon other ones of the factors (953) as discussed above with reference to FIG. 8. Further, the work item may include identifying information (954) for the one or more source storage systems from which to replicate the one or more subsets of the dataset (804) indicated within the one or more work items (956).
Replicating (910), from the respective source storage system, the one or more of the subsets of data corresponding to the quantity of work items (956) may be implemented as described above with reference to FIG. 8. For example, the given session may dequeue the quantity of work items (956), where each work item identifies a portion, or subset, of the dataset (804), and the session may use the information within the work item to initiate a replication of that portion or subset of data (804) from one or more source storage systems identified using the source storage system identifying information (954), as described above with reference to FIG. 8.
In this way, for each of the multiple session, and until each work item is processed, a portion or subset of the dataset (804) is replicated, and when the last work item is processed, and the last portion or subset of the dataset (804) is replicated, the entire dataset (804) will exist on the target storage system (820).
Readers will appreciate that the methods described above may be carried out by any combination of storage systems described above. Furthermore, any of the storage systems described above may also pair with storage that is offered by a cloud services provider such as, for example, Amazon™ Web Services (‘AWS’), Google™ Cloud Platform, Microsoft™ Azure, or others. In such an example, members of a particular pod may therefore include one of the storage systems described above as well as a logical representation of a storage system that consists of storage that is offered by a cloud services provider. Likewise, the members of a particular pod may consist exclusively of logical representations of storage systems that consist of storage that is offered by a cloud services provider. For example, a first member of a pod may be a logical representation of a storage system that consists of storage in a first AWS availability zone while a second member of the pod may be a logical representation of a storage system that consists of storage in a second AWS availability zone.
To facilitate the ability to synchronously replicate a dataset (or other managed objects such as virtual machines) to storage systems that consist of storage that is offered by a cloud services provider, and perform all other functions described in the present application, software modules that carry out various storage system functions may be executed on processing resources that are provided by a cloud services provider. Such software modules may execute, for example, on one or more virtual machines that are supported by the cloud services provider such as a block device Amazon™ Machine Image (AMP) instance. Alternatively, such software modules may alternatively execute in a bare metal environment that is provided by a cloud services provider such as an Amazon™ EC2 bare metal instance that has direct access to hardware. In such an embodiment, the Amazon™ EC2 bare metal instance may be paired with dense flash drives to effectively form a storage system. In either implementation, the software modules would ideally be collocated on cloud resources with other traditional datacenter services such as, for example, virtualization software and services offered by VMware™ such as vSAN™. Readers will appreciate that many other implementations are possible and are within the scope of the present disclosure.
Readers will appreciate that in situations where a dataset or other managed object in a pod is retained in an on-promises storage system and the pod is stretched to include a storage system whose resources are offered by a cloud services provider, the dataset or other managed object may be transferred to the storage system whose resources are offered by a cloud services provider as encrypted data. Such data may be encrypted by the on-promises storage system, such that the data that is stored on resources offered by a cloud services provider is encrypted, but without the cloud services provider having the encryption key. In such a way, data stored in the cloud may be more secure as the cloud has no access to the encryption key. Similarly, network encryption could be used when data is originally written to the on-premises storage system, and encrypted data could be transferred to the cloud such that the cloud continues to have no access to the encryption key.
Through the use of storage systems that consist of storage that is offered by a cloud services provider, disaster recovery may be offered as a service. In such an example, datasets, workloads, other managed objects, and so on may reside on an on-premises storage system and may be synchronously replicated to a storage system whose resources are offered by a cloud services provider. If a disaster does occur to the on-premises storage system, the storage system whose resources are offered by a cloud services provider may take over processing of requests directed to the dataset, assist in migrating the dataset to another storage system, and so on. Likewise, the storage system whose resources are offered by a cloud services provider may serve as an on-demand, secondary storage system that may be used during periods of heavy utilization or as otherwise needed. Readers will appreciate that user interfaces or similar mechanisms may be designed that initiate many of the functions described herein, such that enabling disaster recovery as a service may be as simple as performing a single mouse click.
Through the use of storage systems that consist of storage that is offered by a cloud services provider, high availability may also be offered as a service. In such an example, datasets, workloads, other managed objects, that may reside on an on-premises storage system may be synchronously replicated to a storage system whose resources are offered by a cloud services provider. In such an example, because of dedicated network connectivity to a cloud such as AWS Direct Connect, sub-millisecond latency to AWS from variety of locations can be achieved. Applications can therefore run in a stretched cluster mode without massive expenditures upfront and high availability may be achieved without the need for multiple, distinctly located on-premises storage systems to be purchased, maintained, and so on. Readers will appreciate that user interfaces or similar mechanisms may be designed that initiate many of the functions described herein, such that enabling applications may be scaled into the cloud by performing a single mouse click.
Through the use of storage systems that consist of storage that is offered by a cloud services provider, system restores may also be offered as a service. In such an example, point-in-time copies of datasets, managed objects, and other entities that may reside on an on-premises storage system may be synchronously replicated to a storage system whose resources are offered by a cloud services provider. In such an example, if the need arises to restore a storage system back to a particular point-in-time, the point-in-time copies of datasets and other managed objects that are contained on the storage system whose resources are offered by a cloud services provider may be used to restore a storage system.
Through the use of storage systems that consist of resources that are offered by a cloud services provider, data that is stored on an on-premises storage system may be natively piped into the cloud for use by various cloud services. In such an example, the data that is in its native format as it was stored in the on-premises storage system, may be cloned and converted into a format that is usable for various cloud services. For example, data that is in its native format as it was stored in the on-premises storage system may be cloned and converted into a format that is used by Amazon™ Redshift such that data analysis queries may be performed against the data. Likewise, data that is in its native format as it was stored in the on-premises storage system may be cloned and converted into a format that is used by Amazon™ DynamoDB, Amazon™ Aurora, or some other cloud database service. Because such conversions occurs outside of the on-premises storage system, resources within the on-premises storage system may be preserved and retained for use in servicing I/O operations while cloud resources that can be spun-up as needed will be used to perform the data conversion, which may be particularly valuable in embodiments where the on-premises storage system operates as the primary servicer of I/O operations and the storage systems that consist of resources that are offered by a cloud services provider operates as more of a backup storage system. In fact, because managed objects may be synchronized across storage systems, in embodiments where an on-premises storage system was initially responsible for carrying out the steps required in an extract, transform, load (‘ETL’) pipeline, the components of such a pipeline may be exported to a cloud and run in a cloud environment. Through the use of such techniques, analytics as a service may also be offered, including using point-in-time copies of the dataset (i.e., snapshots) as inputs to analytics services.
Readers will appreciate that applications can run on any of the storage systems described above, and in some embodiments, such applications can run on a primary controller, a secondary controller, or even on both controllers at the same time. Examples of such applications can include applications doing background batched database scans, applications that are doing statistical analysis of run-time data, and so on.
Example embodiments are described largely in the context of a fully functional computer system. Readers of skill in the art will recognize, however, that the present disclosure also may be embodied in a computer program product disposed upon computer readable storage media for use with any suitable data processing system. Such computer readable storage media may be any storage medium for machine-readable information, including magnetic media, optical media, or other suitable media. Examples of such media include magnetic disks in hard drives or diskettes, compact disks for optical drives, magnetic tape, and others as will occur to those of skill in the art. Persons skilled in the art will immediately recognize that any computer system having suitable programming means will be capable of executing the steps of the method as embodied in a computer program product. Persons skilled in the art will recognize also that, although some of the example embodiments described in this specification are oriented to software installed and executing on computer hardware, nevertheless, alternative embodiments implemented as firmware or as hardware are well within the scope of the present disclosure.
Embodiments can include be a system, a method, and/or a computer program product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.
The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.
Computer readable program instructions described herein can be downloaded to respective computing/processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and/or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing/processing device.
Computer readable program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.
Aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to some embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer readable program instructions.
These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function/act specified in the flowchart and/or block diagram block or blocks.
The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions/acts specified in the flowchart and/or block diagram block or blocks.
The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.
Readers will appreciate that the steps described herein may be carried out in a variety of ways and that no particular ordering is required. It will be further understood from the foregoing description that modifications and changes may be made in various embodiments of the present disclosure without departing from its true spirit. The descriptions in this specification are for purposes of illustration only and are not to be construed in a limiting sense. The scope of the present disclosure is limited only by the language of the following claims.

Claims

What is claimed is:

1. A method of asynchronously replicating synchronously replicated data, the method comprising:

determining, at a target storage system, multiple work items corresponding to a dataset stored among multiple source storage systems, wherein each respective work item corresponds to a respective subset of the dataset; and

for each session from among a plurality of sessions operating on the target storage system:

determining one or more computing environment factors affecting performance of replication of data from one or more of the multiple source storage systems to the target storage system;

identifying, for a given session and based on the one or more computing environment factors, a respective source storage system and a quantity of work items; and

replicating, from the respective storage system, one or more subsets of data corresponding to the quantity of work items.

2. The method of claim 1, wherein the dataset is synchronously replicated among the one or more source storage systems.

3. The method of claim 1, wherein at least one of the one or more source storage systems notifies the target storage system that the dataset is ready to be replicated.

4. The method of claim 1, wherein the one or more of the multiple source storage systems are selected based on the one or more computing environment factors, and wherein the one or more source storage systems are less than all of the multiple source storage systems.

5. The method of claim 1, wherein the dataset is a snapshot.

6. The method of claim 1, wherein the target storage system requests metadata describing the dataset, wherein each respective work item is based on the metadata, and wherein the metadata describes a hierarchical structure corresponding to distinct blocks of data that make up the dataset.

7. The method of claim 1, wherein determining the one or more computing environment factors occurs prior to identification of work items to perform.

8. A storage system for asynchronously replicating a synchronously replicated dataset, the storage system comprising a computer processor and a computer memory operatively coupled to the computer processor, the computer memory storing computer program instructions that, when executed by the computer processor, cause the storage system to carry out the steps of:

determine, at a target storage system, multiple work items corresponding to a dataset stored among multiple source storage systems, wherein each respective work item corresponds to a respective subset of the dataset; and

determine one or more computing environment factors affecting performance of replication of data from one or more of the multiple source storage systems to the target storage system;

identify, for a given session and based on the one or more computing environment factors, a respective source storage system and a quantity of work items; and

replicate, from the respective storage system, one or more subsets of data corresponding to the quantity of work items.

9. The storage system of claim 8, wherein the dataset is synchronously replicated among the one or more source storage systems.

10. The storage system of claim 8, wherein at least one of the one or more source storage systems notifies the target storage system that the dataset is ready to be replicated.

11. The storage system of claim 8, wherein the one or more of the multiple source storage systems are selected based on the one or more computing environment factors, and wherein the one or more source storage systems are less than all of the multiple source storage systems.

12. The storage system of claim 8, wherein the dataset is a snapshot.

13. The storage system of claim 8, wherein the target storage system requests metadata describing the dataset, wherein each respective work item is based on the metadata, and wherein the metadata describes a hierarchical structure corresponding to distinct blocks of data that make up the dataset.

14. The storage system of claim 8, wherein determining the one or more computing environment factors occurs prior to identification of work items to perform.

15. An apparatus for asynchronously replicating a synchronously replicated dataset, the apparatus comprising a computer processor and a computer memory operatively coupled to the computer processor, the computer memory storing computer program instructions that, when executed by the computer processor, cause the apparatus to carry out the steps of:

16. The apparatus of claim 15, wherein the dataset is synchronously replicated among the one or more source storage systems.

17. The apparatus of claim 15, wherein at least one of the one or more source storage systems notifies the target storage system that the dataset is ready to be replicated.

18. The apparatus of claim 15, wherein the one or more of the multiple source storage systems are selected based on the one or more computing environment factors, and wherein the one or more source storage systems are less than all of the multiple source storage systems.

19. The apparatus of claim 15, wherein the target storage system requests metadata describing the dataset, wherein each respective work item is based on the metadata, and wherein the metadata describes a hierarchical structure corresponding to distinct blocks of data that make up the dataset.

20. The apparatus of claim 15, wherein determining the one or more computing environment factors occurs prior to identification of work items to perform.