WO2016206198A1

WO2016206198A1 - Storage system

Info

Publication number: WO2016206198A1
Application number: PCT/CN2015/090005
Authority: WO
Inventors: 丁瑞全; 陈国峰; 张家军
Original assignee: 北京百度网讯科技有限公司
Priority date: 2015-06-26
Filing date: 2015-09-18
Publication date: 2016-12-29
Also published as: CN104965677B; CN104965677A

Abstract

Provided is a storage system, comprising a computational resource node, a storage resource node and a PCIe network. The computational resource node and the storage resource node are respectively connected to the PCIe network. The PCIe network, the computational resource node and the storage resource node are physically separated and expandable. The storage system has an improved flexibility, an enhanced access speed of storage resources, and a low cost. Moreover, the storage system can concurrently support disks having at least one of SAS, SATA and PCIe interfaces, and storage media of the disks can comprise an HDD and an SSD. A hybrid storage system is achieved by supporting disks having different interfaces and different storage media. In addition, the storage system can allocate a physical or logical disk to a computational resource node by a dynamic or static configuration, thereby realizing on-demand resource allocation.

Description

Storage System

Cross-reference to related applications

This application claims the priority of the Chinese patent application No. 201510369477.1, which was filed on June 26, 2015 by Beijing Baidu Netcom Technology Co., Ltd. and whose invention is entitled "Storage System".

Technical field

The present invention relates to the field of storage technologies, and in particular, to a storage system.

Background technique

Different applications have different requirements for storage resource capacity, bandwidth, number of read/write operations (IOPS) and reliability per second, which poses a challenge to the design of storage systems. Current storage systems typically have the following implementations: local storage, disk arrays + all-flash arrays, and hybrid disk arrays. The local storage is equipped with storage resources locally on the server. However, due to the different sizes, shapes, and interfaces of different disks, different storage systems need to be designed for different applications. The scalability is poor and cannot be pooled and shared. In the disk array + all-flash array and hybrid disk array solution, mapping or abstraction is required, and the Internet Protocol (IP) storage area network (SAN) or mesh channel (Fibre Channel, FC) is used in the front end. The SAN form provides storage resources externally, and there are deficiencies in terms of flexibility, bandwidth, and cost.

Summary of the invention

The present invention aims to solve at least one of the technical problems in the related art to some extent.

Accordingly, it is an object of the present invention to provide a storage system that can increase flexibility, increase access speed of storage resources, and reduce costs.

To achieve the above objective, a storage system according to an embodiment of the present invention includes: a computing resource node, a storage resource node, and a PCIe network; wherein the computing resource node and the storage resource node are physically separated and respectively connected to the On the PCIe network, and the PCIe network is physically separated from the computing resource node and the storage resource node, and the computing resource node, the storage resource node, and the PCIe network are both expanded.

The storage system provided by the embodiment of the present invention, by physically separating the computing resource node and the storage resource node, are interconnected through independently set PCIe networks, and the components are extensible, which can improve flexibility; directly through the PCIe network The storage resource node is allocated to the computing resource node, which can improve the access speed of the storage resource and reduce the this.

The additional aspects and advantages of the invention will be set forth in part in the description which follows.

DRAWINGS

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from

1 is a schematic structural diagram of a storage system according to an embodiment of the present invention;

2 is a schematic diagram of a PCIe network in an embodiment of the present invention;

3 is a schematic diagram of another PCIe network in the embodiment of the present invention;

4 is a schematic structural diagram of a storage system according to another embodiment of the present invention;

FIG. 5 is a schematic diagram of resource allocation in an embodiment of the present invention; FIG.

6 is a schematic diagram of another resource allocation in an embodiment of the present invention;

7 is a schematic diagram of another resource allocation in an embodiment of the present invention;

FIG. 8 is a schematic diagram of another resource allocation in an embodiment of the present invention; FIG.

FIG. 9 is a schematic diagram of another resource allocation in an embodiment of the present invention; FIG.

FIG. 10 is a schematic diagram of another resource allocation in an embodiment of the present invention.

detailed description

The embodiments of the present invention are described in detail below, and the examples of the embodiments are illustrated in the accompanying drawings, in which the same or similar reference numerals indicate the same or similar modules or modules having the same or similar functions. The embodiments described below with reference to the accompanying drawings are intended to be illustrative of the invention and are not to be construed as limiting. Rather, the invention is to cover all modifications, modifications and equivalents within the spirit and scope of the appended claims.

1 is a schematic structural diagram of a storage system according to an embodiment of the present invention, where the storage system includes:

The computing resource node 11, the storage resource node 12, and the PCIe network 13; PCIe is an abbreviation for PCI-express, and PCI is a Peripheral Component Interconnect (PCI).

The computing resource node 11 and the storage resource node 12 are physically separated and respectively connected to the PCIe network 13, and the PCIe network and the computing resource node and the storage resource node are Physically separate settings, and the computing resource node, the storage resource node and the PCIe network are all extensible.

The number of computing resource nodes may be one or more, and the number of storage resource nodes may be one or more. The computing resource node may specifically be a PCIe Host (PCIe Host).

In the traditional local storage scheme, the central processing unit (CPU) will usually be used. Hard Disk Drive (HDD) and Solid State Disk (SSD) are concentrated in a single physical chassis, and they cannot flexibly expand and change to meet different application requirements.

In this embodiment, by physically separating the computing resource node and the storage resource node, the two are interconnected through the PCIe network. Since the computing resource node and the storage resource node, the PCIe network are independent and scalable, the flexibility can be improved.

The traditional disk array + all-flash array or hybrid disk array, the back end of the SAS, SATA interface HDD and SSD, and the PCIe interface SSD, after abstraction, provide external logical disk access services. The front-end interface is usually an IP SAN or an FC SAN. The export bandwidth is limited, and the high performance of the SSD cannot be fully utilized. IP SAN networks have higher latency and FC SANs have higher costs.

In this embodiment, the storage resource node is directly allocated to the computing resource node through the PCIe network, and there is no additional storage protocol conversion overhead in the middle, and the interconnect bandwidth is very high, which can reduce the network delay, thereby realizing high-speed access of the storage resource, and cut costs. In addition, direct exposure of storage resources to computing resources makes it easier to integrate with existing distributed storage systems. The computing resource node can flexibly use the storage resource node according to its own needs, and utilize the storage resource more efficiently. For example, some storage resources are used as primary storage resources, and some SSDs are used as caches, and cache policies can be defined according to their own needs to truly implement a software-defined storage system.

In another embodiment, the PCIe network includes:

The first level PCIe switch includes: at least one PCIe switch chip and one management module.

Optionally, the PCIe network further includes:

At least one other level PCIe switch, the other level PCIe switch includes: at least one PCIe switch chip;

The other level PCIe switch is connected to the management module; and

The PCIe switch chip in the other-stage PCIe switch is connected to the PCIe switch chip in the first-stage PCIe switch, and/or the PCIe switch chips in different other-stage PCIe switches are connected to each other.

In this embodiment, the PCIe network may be composed of one or more PCIe switches connected according to a certain topological relationship.

For example, referring to FIG. 2, a first-stage PCIe switch may be referred to as a PCIe TOR, and a PCIe TOR may include multiple PCIe switch chips (represented by PCIeX) and a management module (represented by a Mgmt CPU). PCIeX has PCIe switching capability, which can exchange data transmitted between computing resource nodes and storage resource nodes. Mgmt CPU is responsible for configuration management of PCIe network.

For another example, referring to FIG. 3, a PCIe network may further include a multi-level PCIe switch. In a multi-level PCIe switch scenario, there is only one Mgmt CPU in the PCIe network. The Mgmt CPU can be connected to PCIe switch chips in different levels of PCIe switches.

In this embodiment, the PCIe network is constructed by using one or more PCIe switches, which may be different according to services. The need to flexibly build different PCIe networks.

In another embodiment, the storage resource node includes:

A disk having an interface including at least one of the following: Serial Attached SCSI (SAS), Serial Advanced Technology Attachment (SATA), PCIe; wherein SCSI is a small computer system Interface (Small Computer System Interface).

The storage controller has one end connected to the PCIe network and the other end connected to the disk.

When the interface of the disk is SAS or SATA, the form of the disk may be a hard disk drive (HDD) or a solid state disk (SSD). Therefore, the disk may include: SAS HDD, SAS SSD, SATA HDD, SATA SSD.

When the interface of the disk is PCIe, the form of the disk is specifically SSD, so the disk can also be a PCIe SSD.

In addition, the storage controllers can be different depending on the interface of the disk. For example, when the disk is a SAS or SATA (SAS/SATA) interface, the storage controller is a Host Bus Adapter (HBA) or a Redundant Array of Independent Disks (RAID) card. When the storage controller's uplink port (the port connected to the PCIe network) is a PCIe port, the downlink port (the port connected to the disk) includes SAS and/or SATA ports, which can support both SAS and SATA interfaces; when the disk is PCIe In the case of the interface, the storage controller is a PCIe switch chip. In this case, the uplink port of the storage controller is a PCIe port, and the downlink port is also a PCIe port.

It can be understood that there may also be a storage controller, where the uplink interface of the storage controller is a PCIe port, and the downlink port includes at least one of a PCIe port, a SAS port, and a SATA port, when the three types are included at the same time. When the port is used, it can support both the disk of the SAS interface and the disk of the SATA interface and the disk of the PCIe port.

In addition, it can be understood that the storage controller included in the storage system may be one or more types. For example, the storage system includes: a storage controller including a PCIe port, a SAS port, and a SATA port, or The storage system includes: a storage controller that includes a SAS port and/or a SATA port, and a storage controller that is a PCIe port.

In this embodiment, the SAS/SATA interface (the interface may also be referred to as a port) and the PCIe interface are taken as an example. Referring to FIG. 4, the storage resource node may be divided into a SAS/SATA interface resource node and a PCIe interface resource node. In addition, the SAS/SATA interface resource node and the PCIe interface resource node can exist simultaneously under the same PCIe network to support hybrid storage.

The SAS/SATA interface resource node includes: an HBA or a RAID card (HBA/RAID). As a storage controller, one end is connected to the PCIe network, and the other end is connected to the disk. The disk may include at least one of the following items; SAS HDD, SAS SSD, SATA HDD, SATA SSD.

HDD is mainly used for high-capacity storage applications to reduce storage costs. SSD is mainly used for IOPS type with certain requirements. Application to improve performance.

The PCIe interface resource node includes: a PCIe switch, which is a storage controller, one end is connected to the PCIe network, and the other end is connected to the disk, and the disk includes: PCIe SSD.

PCIe SSD, with extremely high IOPS, can significantly improve the business performance of IOPS applications, such as databases.

In this embodiment, by connecting storage resource nodes of different interface types and/or different storage media to the PCIe network, the disks of the SAS, SATA, and PCIe interfaces can be supported under the same PCIe network, and the storage medium of the disk can be Including HDD and SSD (for example, support for HDD and SSD in SAS or SATA interface, SSD support in PCIe interface), therefore, under the same PCIe network, SAS HDD, SAS SSD, SATA HDD, SATA SSD, PCIe SSD can Any combination of hybrid storage systems to support high-capacity storage applications to reduce costs, high-bandwidth, high-IOPS applications to improve business performance, and even support high-capacity, low-cost, high-bandwidth, and high IOPS Demand.

In another embodiment, the PCIe network is further configured to:

Allocating the storage resource node to the computing resource node in the form of a physical disk or a logical disk, wherein a single physical disk or a logical disk is allocated to a single computing resource node, or a single physical disk or a logical disk is simultaneously allocated to Multiple different computing resource nodes.

Specifically, the PCIe network includes a management module (Mgmt CPU), when the disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing resource node in a physical disk format, and a single physical disk is allocated to When a single computing resource node is used, the management module is used to:

Configure the correspondence between each compute resource node and each PCIe SSD that is granular to the physical disk.

or,

The PCIe network includes a management module (Mgmt CPU), when the disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing resource node in a logical disk form, and a single logical disk is allocated to a single computing The resource node, the PCIe SSD includes a PCIe SSD controller supporting SR-IOV function.

The PCIe SSD controller is configured to generate a VF, and divide the PCIe SSD into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where different VFs correspond to different logical blocks. ;

The management module is configured to configure a correspondence between each computing resource node and each VF.

or,

The PCIe network includes a management module (Mgmt CPU), when the disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing resource node in a logical disk format, and a single logical disk is simultaneously allocated to multiple The PCIe SSD includes a PCIe SSD controller supporting SR-IOV function when different computing resource nodes are used.

The PCIe SSD controller is configured to generate a VF, and divide the PCIe SSD into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where at least one logical block corresponds to multiple VFs. ;

or,

The PCIe network includes a management module (Mgmt CPU), and the disk of the storage resource node is a disk of a SAS or SATA interface, and the disk of the SAS or SATA interface is allocated to a computing resource node in a physical disk format, and a single When the physical disk is allocated to a single computing resource node, the storage resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or the RAID controller is configured to generate a VF, and establish a mapping relationship between the disk of the SAS or SATA interface with a granularity of the physical disk and the VF, where different VFs correspond to different physical disks;

or,

The PCIe network includes a management module (Mgmt CPU), and the disk of the storage resource node is a disk of a SAS or SATA interface, and the disk of the SAS or SATA interface is allocated to a computing resource node in a logical disk format, and a single When the logical disk is allocated to a single computing resource node, the storage resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or RAID controller is configured to generate a VF, and divide the disk of the SAS or SATA interface into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where different VFs are used. Corresponding to different logic blocks;

or,

The PCIe network includes a management module (Mgmt CPU), and the disk of the storage resource node is a disk of a SAS or SATA interface, and the disk of the SAS or SATA interface is allocated to a computing resource node in a logical disk format, and a single When the logical disk is simultaneously allocated to a plurality of different computing resource nodes, the storage resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or RAID controller is configured to generate a VF, and divide the disk of the SAS or SATA interface into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where at least one logic The block corresponds to multiple VFs;

For example, a PCIe SSD can be supported as a physical disk to be allocated to a computing resource node (such as a PCIe host) on demand. Within the PCIe network, any PCIe SSD is a separate PCIe device. The Mgmt CPU is responsible for the scanning and discovery of PCIe devices and PCIe hosts in the PCIe network, and configures the routing table of the PCIe network to statically or dynamically allocate specific PCIe devices to specific PCIe hosts according to the requirements of the PCIe host.

In this embodiment, referring to FIG. 5, there are four PCIe SSDs in the PCIe network. After being configured by the Mgmt CPU, the PCIe is SSD A is assigned to computing resource node A, and PCIe SSD B, PCIe SSD C, and PCIe SSD D are assigned to computing resource node B. In addition, through the configuration of the Mgmt CPU, the complex PCIe physical network can be simplified into a logical PCIe Bridge, and the computing resource node can only see the PCIe Bridge, thus shielding the influence of the physical topology change on the computing resource node.

As another example, a PCIe SSD can be supported to be allocated to a computing resource node as needed in a logical disk format. For a large-capacity PCIe SSD, if the entire disk can only be allocated to a certain computing resource node, the capacity may exceed its demand, resulting in low resource utilization and ultimately cost waste. In this embodiment, the PCIe SSD can be divided into multiple logical blocks, and then the logical blocks are allocated to different computing resource nodes, so that the management and allocation of resources can be performed with a smaller granularity to improve resource utilization.

In this embodiment, referring to FIG. 6, the physical disk PCIe SSD A is divided into examples, and the logical blocks after the splitting are referred to as SSD block A, SSD block B and SSD block C. A physical presence controller (PCIe SSD Controller) is provided in each PCIe SSD. When the controller supports single root I/O virtualization (SR-IOV), it can be logically formed. Multiple Virtual Functions (VF), each VF is a separate PCIe device in the PCIe network.

The PCIe SSD Controller can map logical blocks to different VFs. For example, referring to Figure 6, SSD block A is mapped to VF-1, and SSD block B and SSD block C are mapped to VF-2. The Mgmt CPU is responsible for allocating different VFs to different computing resource nodes (the same VF cannot be assigned to multiple computing resource nodes). For example, referring to Figure 6, assigning VF-1 to computing resource node A, assigning VF-2 Give computing resource Node B. Therefore, the computing resource node A can access the SSD block A, and the computing resource node B can access the SSD block B and the SSD block C, so that the PCIe SSD is allocated to the computing resource node in a logical disk form as needed.

As another example, multiple computing resource nodes can be simultaneously accessed to access the same PCIe SSD logic block. The PCIe SSD Controller can map the same SSD logic block to different VFs, and the Mgmt CPU is responsible for allocating VFs to different computing resource nodes. Therefore, different computing resource nodes can access the same PCIe SSD logic block at the same time to achieve data sharing. Multiple computing resource nodes can simultaneously read the same PCIe SSD logic block. Technically, the same PCIe SSD logic block can be written at the same time, but the consistency of the data requires the upper layer software to coordinate and guarantee.

In this embodiment, referring to FIG. 7, the PCIe SSD Controller maps the SSD block A and the SSD block B to the VF-1, the SSD block B and the SSD block C to the VF-2, and the Mgmt CPU allocates the VF-1 to the computing resource. The node A allocates the VF-2 to the computing resource node B, so that the computing resource node A and the computing resource node B can simultaneously access the SSD block B, thereby supporting multiple computing resource nodes to simultaneously access the same PCIe SSD logical block.

Of course, it can be understood that when a physical disk of a PCIe SSD is divided into a logical block, multiple computing resource nodes can simultaneously access the same physical disk.

As another example, a SAS/SATA interface disk can be supported as a physical disk to be allocated to a computing resource node as needed. HBA/RAID can include HBA/RAID Controller in hardware. When HBA/RAID Controller does not support SR-IOV, it can only be managed as a PCIe device by Mgmt CPU. The disk on the back end is invisible to PCIe network. of. Therefore, all the disks connected to a certain HBA/RAID Controller can be allocated to a computing resource node as a whole. The granularity of resource allocation is large, and it is difficult to achieve efficient use of resources.

In this embodiment, the application is implemented in an scenario where the HBA/RAID Controller supports SR-IOV. When the HBA/RAID Controller supports SR-IOV, it supports mapping different disks to different VFs. Each VF is a separate PCIe device in the PCIe network, and the Mgmt CPU is responsible for allocating VFs to different computing resource nodes. The same VF cannot be assigned to multiple compute resource nodes. Therefore, it is possible to indirectly implement different physical disks to be allocated to different computing resource nodes.

For example, referring to Figure 8, the HBA/RAID Controller maps Disk-1 and Disk 2 to VF-1, Disk-3 and Disk-4 to VF-2, and the Mgmt CPU assigns VF-1 to Computing Resource Node A. The VF-2 is allocated to the computing resource node B, so that the computing resource node A can access the disk-1 and the disk-2, and the computing resource node B can access the disk-3 and the disk-4, thereby supporting the SAS/SATA interface disk. It is allocated to the computing resource node as needed in the form of a physical disk.

As another example, a SAS/SATA interface disk can be supported as a logical disk to be allocated to a computing resource node as needed. The HBA/RAID Controller can aggregate one or more physical disks, divide it into one or more logical disks, and then map the logical disks to different VFs. Each VF is a separate PCIe device in the PCIe network, and the Mgmt CPU is responsible for allocating VFs to different compute nodes. Therefore, the management allocation of resources can be performed at a smaller granularity to improve resource utilization.

In this embodiment, referring to FIG. 9, the HBA/RAID Controller maps logical disk-1 and logical disk-2 to VF-1, logical disk-3 and logical disk-4 to VF-2, and Mgmt CPU will VF- 1 is allocated to the computing resource node A, and VF-2 is allocated to the computing resource node B, so that the computing resource node A can access the logical disk-1 and the logical disk-2, and the computing resource node B can access the logical disk-3 and the logic Disk-4, which supports the allocation of SAS/SATA interface disks to compute resource nodes as logical disks.

As another example, multiple computing resource nodes can be simultaneously accessed to access the same logical SAS/SATA interface disk. The HBA/RAID Controller can map the same logical disk to different VFs, and the Mgmt CPU is responsible for allocating VFs to different computing resource nodes. Therefore, different computing resource nodes can access the same logical disk at the same time to achieve data sharing. Multiple computing resource nodes can simultaneously read the same logical disk. Technically, the same logical disk can be written at the same time, but the consistency of the data needs to be coordinated by the upper layer software.

In this embodiment, referring to FIG. 10, the HBA/RAID Controller maps logical disk-1 and logical disk-2 to VF-1, logical disk-2 and logical disk-3 to VF-2, and Mgmt CPU will VF- 1 is allocated to the computing resource node A, and the VF-2 is allocated to the computing resource node B, so that both the computing resource node A and the computing resource node B can access the logical disk-2, thereby supporting multiple computing resource nodes to simultaneously access the same Logical SAS/SATA interface disk.

In the foregoing embodiment of the resource allocation, a physical disk or a logical disk may be allocated to the computing resource node by using a dynamic or static configuration, and different numbers and different types of storage resources may be configured according to the requirements of the computing resource node, which are flexible and changeable, and can satisfy each A variety of business needs. The number of storage resources allocated to the computing resource node can be dynamically increased or decreased. When the service demand increases, the number of storage resources (such as PCIe SSD) can be increased to meet the peak demand; when the service demand decreases, the number of PCIe SSDs can be reduced. It is allocated to other computing resource nodes to improve resource utilization and reduce overall system cost. Especially suitable for public cloud platforms, it can flexibly build servers with different configurations, and can support both large-capacity storage applications and high IOPS applications in the same platform, and even support applications with storage capacity and IOPS at the same time. To meet the needs of differentiated and changeable public cloud users.

It should be noted that in the description of the present invention, the terms "first", "second" and the like are used for descriptive purposes only, and are not to be construed as indicating or implying relative importance. Further, in the description of the present invention, the meaning of "a plurality" means at least two unless otherwise stated.

Any process or method description in the flowcharts or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a particular logical function or process. And the scope of the preferred embodiments of the invention includes additional implementations, in which the functions may be performed in a substantially simultaneous manner or in an opposite order depending on the functions involved, in the order shown or discussed. It will be understood by those skilled in the art to which the embodiments of the present invention pertain.

It should be understood that portions of the invention may be implemented in hardware, software, firmware or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques well known in the art: having logic gates for implementing logic functions on data signals. Discrete logic circuits, application specific integrated circuits with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

One of ordinary skill in the art can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by a program to instruct related hardware, and the program can be stored in a computer readable storage medium. When executed, one or a combination of the steps of the method embodiments is included.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. The integrated modules, if implemented in the form of software functional modules and sold or used as stand-alone products, may also be stored in a computer readable storage medium.

The above mentioned storage medium may be a read only memory, a magnetic disk or an optical disk or the like.

In the description of the present specification, the terms "one embodiment", "some embodiments", "example", "specific examples", The description of "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiments or examples are included in at least one embodiment or example of the invention. In the present specification, the schematic representation of the above terms does not necessarily mean the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples.

Although the embodiments of the present invention have been shown and described, it is understood that the above-described embodiments are illustrative and are not to be construed as limiting the scope of the invention. The embodiments are subject to variations, modifications, substitutions and variations.

Claims

A storage system, comprising:

Computing resource nodes, storage resource nodes, and PCIe networks;

The computing resource node and the storage resource node are physically separated and respectively connected to the PCIe network, and the PCIe network is physically separated from the computing resource node and the storage resource node. The computing resource node, the storage resource node, and the PCIe network are all scalable.
The system of claim 1 wherein said PCIe network comprises:

The first level PCIe switch includes: at least one PCIe switch chip and one management module.
The system of claim 2, wherein the PCIe network further comprises:

At least one other level PCIe switch, the other level PCIe switch includes: at least one PCIe switch chip;

The other level PCIe switch is connected to the management module; and

The PCIe switch chip in the other-stage PCIe switch is connected to the PCIe switch chip in the first-stage PCIe switch, and/or the PCIe switch chips in different other-stage PCIe switches are connected to each other.
The system according to any one of claims 1 to 3, wherein the storage resource node comprises:

a disk having an interface including at least one of the following: SAS, SATA, PCIe;

The storage controller has one end connected to the PCIe network and the other end connected to the disk.
The system according to claim 4, wherein when the disk is a SAS or SATA interface, the storage controller is an HBA or a RAID card, and the disk includes at least one of the following: a SAS HDD, SAS SSD, SATA HDD, SATA SSD.
The system according to claim 4 or 5, wherein when the disk is a PCIe interface, the storage controller is a PCIe switch chip, and the disk is a PCIe SSD.
The system according to any one of claims 1 to 6, wherein the PCIe network is further configured to:

Allocating the storage resource node to the computing resource node in the form of a physical disk or a logical disk, wherein a single physical disk or a logical disk is allocated to a single computing resource node, or a single physical disk or a logical disk is simultaneously allocated to Multiple different computing resource nodes.
The system according to claim 7, wherein the PCIe network comprises a management module, wherein a disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing resource node in a physical disk format, and When a single physical disk is assigned to a single computing resource node, the management module is used to:

Configure the correspondence between each compute resource node and each PCIe SSD that is granular to the physical disk.
The system according to claim 7 or 8, wherein the PCIe network comprises a management module, wherein a disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing resource node in a logical disk format, And when a single logical disk is allocated to a single computing resource node, the PCIe SSD includes a PCIe SSD controller supporting SR-IOV function,

The PCIe SSD controller is configured to generate a VF, and divide the PCIe SSD into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where different VFs correspond to different logical blocks. ;

The management module is configured to configure a correspondence between each computing resource node and each VF.
The system according to any one of claims 7-9, wherein the PCIe network comprises a management module, wherein a disk of the storage resource node is a PCIe SSD, and the PCIe SSD is allocated to a computing in the form of a logical disk. a resource node, and when a single logical disk is simultaneously allocated to a plurality of different computing resource nodes, the PCIe SSD includes a PCIe SSD controller supporting SR-IOV function,

The PCIe SSD controller is configured to generate a VF, and divide the PCIe SSD into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where at least one logical block corresponds to multiple VFs. ;

The management module is configured to configure a correspondence between each computing resource node and each VF.
The system according to any one of claims 7 to 10, wherein the PCIe network comprises a management module, when the disk of the storage resource node is a disk of a SAS or SATA interface, and the SAS or SATA interface The disk is allocated to the computing resource node in the form of a physical disk, and when a single physical disk is allocated to a single computing resource node, the storage resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or the RAID controller is configured to generate a VF, and establish a mapping relationship between the disk of the SAS or SATA interface with a granularity of the physical disk and the VF, where different VFs correspond to different physical disks;

The management module is configured to configure a correspondence between each computing resource node and each VF.
The system according to any one of claims 7 to 11, wherein the PCIe network includes a management module, when the disk of the storage resource node is a disk of a SAS or SATA interface, and the SAS or SATA interface The disk is allocated to the computing resource node in the form of a logical disk, and when a single logical disk is allocated to a single computing resource node, the storage resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or RAID controller is configured to generate a VF, and divide the disk of the SAS or SATA interface into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where different VFs are used. Corresponding to different logic blocks;

The management module is configured to configure a correspondence between each computing resource node and each VF.
The system according to any one of claims 7 to 12, wherein the PCIe network comprises a management module, wherein a disk of the storage resource node is a disk of a SAS or SATA interface, and the SAS or SATA interface The disk is allocated to the computing resource node in the form of a logical disk, and a single logical disk is simultaneously allocated to a plurality of different computing resources. The source resource node further includes an HBA or a RAID controller supporting the SR-IOV function.

The HBA or RAID controller is configured to generate a VF, and divide the disk of the SAS or SATA interface into one or more logical blocks, and establish a mapping relationship between the logical block and the VF, where at least one logic The block corresponds to multiple VFs;

The management module is configured to configure a correspondence between each computing resource node and each VF.