WO2017223265A1 - Synchronisation de niveau de segment d'une mémoire de données en nuage et de systèmes de fichiers locaux - Google Patents

Synchronisation de niveau de segment d'une mémoire de données en nuage et de systèmes de fichiers locaux Download PDF

Info

Publication number
WO2017223265A1
WO2017223265A1 PCT/US2017/038661 US2017038661W WO2017223265A1 WO 2017223265 A1 WO2017223265 A1 WO 2017223265A1 US 2017038661 W US2017038661 W US 2017038661W WO 2017223265 A1 WO2017223265 A1 WO 2017223265A1
Authority
WO
WIPO (PCT)
Prior art keywords
shard
server
local
version
cloud
Prior art date
Application number
PCT/US2017/038661
Other languages
English (en)
Inventor
David M. Shaw
Matthew M. Mcdonald
Russell A. Neufeld
Christopher S. Lacasse
Original Assignee
Nasuni Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nasuni Corporation filed Critical Nasuni Corporation
Publication of WO2017223265A1 publication Critical patent/WO2017223265A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10—File systems; File servers
    • G06F16/17—Details of further file system functions
    • G06F16/178—Techniques for file synchronisation in file systems

Definitions

  • the present application relates generally to data storage, and more particularly to synchronizing data stored in a cloud based network-attached file system at the data shard level or sub-directory level.
  • the versioned file system comprises a set of structured data representations, such as XML.
  • the interface creates and exports to a data store a first structured data representation corresponding to a first version of the local file system.
  • the first structured data representation is an XML tree having a root element, a single directory (the "root directory") under the root element, zero or more directory elements associated with the root directory, and zero or more elements (such as files) associated with a given directory element.
  • Each directory in turn can contain zero or more directories and zero or more files.
  • the interface Upon a change within the file system (e.g., file creation, file deletion, file modification, directory creation, directory deletion and directory modification), the interface creates and exports a second structured data representation corresponding to a second version of the file system.
  • the second structured data representation differs from the first structured data representation up to and including the root element of the second structured data representation.
  • the second structured data representation differs from the first structured data representation in one or more (but not necessarily all) parent elements with respect to the structured data element in which the change within the file system occurred.
  • the interface continues to generate and export structured data representations to the data store, preferably at given "snapshot" times when changes within the file system have occurred.
  • the data store comprises any type of back-end storage device, system or architecture.
  • the data store comprises one or more cloud storage service providers.
  • a given structured data representation is then used to retrieve an associated version of the file system.
  • the versioned file system only requires write-once behavior from the data store to preserve its complete state at any point-in-time.
  • a problem with the above system is that a change to any file or directory in the file system causes a new version of each parent directory all the way up to the root. This causes additional processing time and resources to create each new "version" of the file system. Also, to determine what file or directory has changed between versions of the file system, the entire directory structure needs to be "walked.” In a large file system with a large user base, the processing overhead required to maintain this directory structure is significant. It would be desirable to create versions of a more granular portion of a file system without having to create a snapshot of the entire file system.
  • Figure 1 is a block diagram illustrating how a shared versioned file system interfaces a local version of the shared versioned file system to an object-based data store;
  • Figure is a block diagram of a representation implementation of a portion of the interface shown in Figure 1;
  • Figure 3 is a more detailed implementation of the interface where there are a number of local versions of the shared versioned file system of different types;
  • Figure 4 illustrates a filer server implemented as an appliance within a local processing environment
  • Figure 5 is a block diagram of the architecture of a shared versioned file system according to an embodiment
  • Figure 6 illustrates the portion of the tree (as shown in Figure 5) after a change to the contents of a file has occurred in the local version of the shared versioned file system;
  • Figure 7 is a block diagram of a system for running a shared versioned file system according to an embodiment
  • Figure 8 is a flow chart of a method for sending updated data to cloud storage according to an embodiment
  • Figure 9 illustrates a table of updates to a shared versioned file system maintained by the filer server
  • Figure 10 is a flow chart of a method for synchronizing updates from a local version of a shared versioned file system to a cloud data store according to an embodiment
  • Figure 11 is a flow chart of a method for synchronizing updates from a cloud- based a shared versioned file system to a local version of same in an embodiment
  • Figure 12 is a simplified illustration of a representative shard in an embodiment
  • Figure 13 is a simplified illustration of the representative shard from Figure 12 after an update in one embodiment
  • Figure 14 is a simplified illustration of a representative directory entry in the representative shard of Figure 13.
  • Figure 15 is a Table describing filer server operations that depend on a state of a directory entry.
  • Figure 1 illustrates a simplified system 10 for providing a shared versioned file system.
  • the system 10 includes local versions ioo, 101 of the shared versioned file system and an object-based data store 120.
  • the object-based store 120 can be a "write-once" store and may comprise a "cloud” of one or more storage service providers.
  • Each interface or filer server 110, 111 exposes a respective local version ioo, 101 of a "shared versioned file system” that only requires write-once behavior from the object-based data store 120 to preserve substantially its "complete" state at any point-in-time.
  • each filer server ioo, 101 provides for a local version ioo, 101 of the shared versioned file system that has complete data integrity to the cloud. In particular, this solution circumvents the problem of a lack of reliable atomic object replacement in cloud-based object repositories.
  • the filer servers ioo, 101 are not limited for use with a particular type of back-end data store.
  • the filer servers ioo, 101 When the filer servers ioo, 101 are positioned in "front" of data store 120, the filer servers ioo, 101 have the effect of turning whatever is behind it into respective local versions of a "shared versioned file system" ("SVFS").
  • the SVFS is a construct that is distinct from the filer server itself, and the SVFS continues to exist irrespective of the state or status of the filer server (from which it may have been generated).
  • the SVFS is self-describing, and it can be accessed and managed separately from the back-end data store, or as a component of that data store.
  • the SVFS (comprising a set of structured data representations) is location-independent.
  • the SVFS resides within a single storage service provider (SSP) although, as noted above, this is not a limitation.
  • SSP single storage service provider
  • a first portion of the SVFS resides in a first SSP, which a second portion resides in a second SSP.
  • any given SVFS portion may reside in any given data store (regardless of type), and multiple VFS portions may reside across multiple data store(s).
  • the SVFS may reside in an "internal" storage cloud (i.e. a storage system internal to an enterprise), an external storage cloud, or some combination thereof.
  • the interface or filer server 104 can be implemented as a machine.
  • the filer server 104 is a rack-mounted server appliance comprising of hardware and software.
  • the hardware typically includes one or more processors that execute software in the form of program instructions that are otherwise stored in computer memory to comprise a "special purpose" machine for carrying out the
  • the filer server 104 is implemented as a virtual machine or appliance (e.g., via VMware ® , or the like), as software executing on a server, or as software executing on the native hardware resources of the local version of the SVFS.
  • the filer server 104 serves to transform the data representing the local version of the SVFS (a physical construct) into another form, namely, a shared versioned file system comprising a series of structured data representations that are useful to reconstruct the shared versioned file system to any point-in-time.
  • XML extensible markup language
  • An XML document typically contains a single root element (or a root element that points to one or more other root elements). Each element has a name, a set of attributes, and a value consisting of character data, and a set of child elements.
  • the interpretation of the information conveyed in an element is derived by evaluating its name, attributes, value, and position in the document.
  • the filer server 104 generates and exports to the write-once data store a series of structured data representations (e.g., XML docu ments) and data objects that together comprise the shared versioned file system.
  • the structured data representations are stored in the data store 120.
  • the XM L representations are encrypted before export to the data store.
  • the transport may be performed using known techniques.
  • REST Representational State Transfer
  • SOAP Simple Object Access Protocol
  • XM L-based messages are exchanged over a computer network, normally using HTTP (Hypertext Transfer Protocol) or the like.
  • Transport layer security mechanisms such as HTTP over TLS (Transport Layer Security) may be used to secu re messages between two adjacent nodes.
  • An XML document and/or a given element or object therein is addressable via a Uniform Resource Identifier (URI). Familiarity with these technologies and standards is presu med.
  • URI Uniform Resource Identifier
  • FIG. 2 is a block diagram of a representative implementation of how the interface or filer server 110/111 captures all (or given) read/write events from a local version of shared versioned file system 200.
  • the interface comprises a file system agent (FSA) 202 that is positioned within a data path between a local version of shared versioned file system 200 and its local storage 206.
  • the file system agent 202 has the capability of "seeing" all (or some configurable set of) read/write events output from the local file system.
  • the interface/filer server also comprises a content control service (CCS) 204 as will be described in more detail below.
  • the content control service is used to control the behavior of the file system agent.
  • the object-based data store is represented by the arrows directed to "storage" which, as noted above, typically comprises any back-end data store including, without limitation, one or more storage service providers.
  • the local version of the shared versioned file system stores local user files (the data) in their native form in cache 208.
  • Reference numeral 210 represents that portion of the cache that stores pieces of metadata (the structured data representations, as will be described) that are exported to the back-end data store (e.g., the cloud).
  • Figure 3 is a block diagram illustrating how the interface may be used with different types of local file system architectures. I n particular, Figure 3 shows the CCS (in this drawing a Web-based portal) controlling three (3) FSA instances.
  • the file system agent 306 is used with three (3) different local versions of the shared versioned file system: NTFS 300 executing on a Windows operating system platform 308, MacFS (also referred to as "HFS+" (HFSPIus)) 302 executing on an OS X operating system platform 310, and EXT3 or XFS 304 executing on a Linux operating system platform 312.
  • each platform may be controlled from a single CCS instance 314, and one or more external storage service providers may be used as an external object repository 316.
  • external object repository 316 there is no requirement that multiple SSPs be used, or that the data store be provided using an SSP.
  • FIG. 4 illustrates the interface/filer server implemented as an appliance within a local processing environment.
  • the version of 400 for the local version of the shared versioned file system is received over Ethernet and represented by the arrow identified as "NAS traffic.” That traffic is provided to smbd layer 402, which is a SAMBA file server daemon that provides CI FS (Window-based) file sharing services to clients.
  • the layer 402 is managed by the operating system kernel 404 is the usual manner.
  • the local version of the shared versioned file system is represented (in this example) by the FUSE kernel module 406 (which is part of the Linux kernel distribution). Components 400, 402 and 404 are not required to be part of the appliance.
  • the file transfer agent 408 of the interface is associated with the FUSE module 406 as shown to intercept the read/write events as described above.
  • the CCS (as described above) is implemented by a pair of modules (which may be a single module), namely, a cache manager 410, and a volume manager 412. Although not shown in detail, preferably there is one file transfer agent instance 408 for each volume of the local file system.
  • the cache manager 410 is responsible for management of "chunks" with respect to a local disk cache 414. This enables the interface or filer server described herein to maintain a local cache of the data structures (the structured data representations) that comprise the shared versioned file system.
  • the volu me manager 412 maps the root of the FSA data to the cloud (as will be described below), and it further understands the one or more policies of the cloud storage service providers.
  • the volume manager also provides the application programming interface (API) to these one or more providers and communicates the structured data representations (that comprise the shared versioned file system) through a transport mechanism 416 such as cURL.
  • cURL is a library and command line tool for transferring files with URL syntax that supports various protocols such as FTP, FTPS, HTTP, HTTPS, SCP, SFTP, TFTP, TELNET, DICT, LDAP, LDAPS and FILE.
  • cURL also supports SSL certificates, HTTP POST, HTTP PUT, FTP uploading, HTTP form based u pload, proxies, cookies, user + password authentication, file transfer resume, proxy tunneling, and the like.
  • the structured data representations preferably are encrypted and compressed prior to transport by the transformation module 418.
  • the module 418 may provide one or more other data transformation services, such as duplicate elimination.
  • the encryption, compression, duplicate elimination and the like, or any one of such functions, are optional.
  • a messaging layer 420 e.g., local socket-based IPC
  • IPC may be used to pass messages between the file system agent instances, the cache manager and the volume manager. Any other type of message transport may be used as well.
  • the interface/filer server shown in Figure 4 may be implemented as a standalone system, or as a managed service. In the latter case, the system executes in an end user (local file system) environment.
  • a managed service provider provides the system (and the versioned file system service), preferably on a fee or subscription basis, and the data store (the cloud) typically is provided by one or more third party service providers.
  • the shared versioned file system may have its own associated object-based data store, but this is not a requirement, as its main operation is to generate and manage the structured data representations that comprise the shared versioned file system.
  • the cloud preferably is used just to store the structured data representations, preferably in a write-once manner, although the "shared versioned file system" as described herein may be used with any back-end data store and can be a write-many data store.
  • the file system agent 408 is capable of completely recovering from the cloud (or other store) the state of the local version of the shared versioned file system and providing immediate file system access (once FSA metadata is recovered).
  • the FSA can also recover to any point-in-time for the whole shared versioned file system, a directory and all its contents, a portion of a directory (e.g., a shard) and it contents, a single file, or a piece of a file.
  • Figure 5 is a block diagram of the architecture of a shared versioned file system
  • the architecture 50 includes a root-level directory 500 and first-level directories 500-1, 500-2.
  • First level directory 500-2 includes sub-directory 2-1 502, which is divided into shards 1, 2, and 3 (corresponding to reference numbers 503-1, 503-2, 503- 3) (in general, shard 503).
  • Each shard 503 is a portion of sub-directory 2-1 502.
  • files 1, 2, and 3 (corresponding to reference numbers 504-1, 504-2, and 504-3) in sub-directory 2-1 502 are assigned to shard 1 503-1.
  • Shard 2 503-2 and shard 3 503-3 can also include files and/or metadata that belong to sub-directory 2-1 502.
  • Each file 504 is divided into one more chunks, such as chunks 1, 2, 3
  • di rectory /sub-di rectory, file, and chunk of shared versioned file system 50 can be represented by an inode.
  • Example inode numbers for the following components of shared versioned file system 50 are illustrated in parentheticals: sub-directory 2-1 502 (10), file 1 504-1 (101), file 2 504-2 (102), file 3 504-3 (103), and chunk 1 505-1 (1001). Additional inode numbers are illustrated in Figure 5.
  • Shard 503 can have an arbitrary number of files and/or metadata from subdirectory 2-1 502. I n addition, or in the alternative, shard 503 can have a maximum number of files and/or metadata, for example to provide an increased size (horizontally and/or vertically) of the shared versioned file system.
  • Each shard 503 has a manifest that identifies the files (by inode number) assigned to that shard.
  • manifest 540 of shard 1 503-1 identifies inodes 101, 102, and 103.
  • the manifest 540 also includes metadata about each inode, such as the version of the shard in which the inode (file) was created and the version of the shard in which the inode (file) was last modified.
  • the manifest can also include a flag or bit to indicate whether any component of the shard has been modified, including the manifest itself.
  • each file 504 has a manifest that identifies the chunks (by inode nu mber) that make up the data of the file.
  • manifest 550 of file 2 504-2 identifies inodes 1001, 1002, and 1003.
  • the manifest also includes metadata about each inode, such as the relationship or offset between each inode.
  • the manifest can also include a flag or bit to indicate whether any component of the file has been modified, including the manifest itself.
  • Figure 6 is a block diagram of the architecture of the shared versioned file system 50 after a change to chunk 1 505-1 in file 2 504-2.
  • the change to chunk 1* 505-1 propagates to file 2* 504-2.
  • the modification to chunk 1* 505-1 causes file 2* 504-2 to appear as modified or "dirty.”
  • the modified or dirty file 2* 504-2 causes shard 1* 503-1 to appear as modified, which in turn causes sub-directory 2-1* 502 to appear as modified.
  • the modification to chunk 1* 505-1 does not propagate past sub-directory 2-1* 502, such as to directory 2 501-2 or root 500.
  • directory 2 501-2 and root 500 appear as unmod ified even if su b-directory 2-1* 502 appears as modified.
  • a change to any portion of the shared versioned file system 50 only propagates to the closest directory or sub-directory level. For example, a change to shard 2 503-2 propagates to sub-directory 2-1 502 but not to directory 2 501-2 or root 500. Similarly, a change to sub-directory 2-1 propagates to directory 2 501-2 but not to root 500. [0036] By limiting the propagation of change events to the closest directory or subdirectory, shared versioned file system 50 can be synchronized more efficiently across local interfaces running respective local versions of the shared versioned file system.
  • a modification to a file or shard causes an update flag in the respective manifest to turn on, which makes the corresponding file or shard appear as modified.
  • the modification to chunk 1* 505-1 automatically causes the update flag in manifest* 550 to tu rn on, which in turn causes file 2* 504-2 to appear as mod ified.
  • the modification to file 2* 504-2 causes the u pdate flag in manifest* 540 to turn on, which in turn causes shard 1* 503-1 to appear as modified.
  • sub-directory 2-1* 502 also appears as mod ified since shard 1* 503-1 is a portion of sub-directory 2-1* 502.
  • FIG. 7 is a block diagram of a system 70 for running a shared versioned file system according to an embodiment.
  • the system 70 includes operations server 700, filer servers 710, 720, and user computers 712, 714, 722, 724.
  • Filer servers 710, 720 can be the same as FSA 202 or FSA 306 described above.
  • Each filer server provides a respective local version of the shared versioned file system to its respective user computers.
  • Filer server 710 exposes local version A 730 of the shared versioned file system to local computers 712, 714.
  • Filer server 720 exposes local version B 740 of the shared versioned file system to local computers 722, 724.
  • Local version A 730 and local version B 740 can represent the same or different versions of the shared versioned file system based on how recently the respective filer server 710, 720 have retrieved updates to the shared versioned file system from operations server 700 and cloud storage 750. If filer servers 710, 720 have retrieved updates to the shared versioned file system up to the same change event (as discussed below), local versions 730, 740 of the shared versioned file system are identical.
  • the filer servers 710, 720 can communicate with respective user computers over a network protocol such as Network File System (NFS), Server Message Block (SMB), or Common Internet File System (CIFS).
  • the operations server 700 is a NASUNI ® Operations Center, available from Nasuni Corporation of Massachusetts.
  • a user on user computer 712 makes a modification to a document that corresponds to file 2 504-2 (using the example of Figures 5 and 6, discussed above).
  • the modification occurs in the portion of file 2 504-2 corresponding to chunk 1 505-1.
  • Filer server7io saves a new version of File 2 504-2 locally.
  • the new version of file 2 504-2 includes modified manifest 540* that contains modified chunk 1* 505-1 and pointers to unmodified chunk 2 505-2 and unmodified chunk n 505-n.
  • Filer server 710 also saves a new version of shard 1 503-1 locally.
  • the new version (e.g., version 2) of shard 1 503-1 (i.e., modified shard 1* 503-1) includes a new manifest 550* that includes the the inode numbers of each file in modified shard 1* 503-1.
  • manifest 550* includes inodes
  • manifest 550* indicates that inode 102 was last modified in version 2 of shard 1 (i.e., modified shard 1* 503-1). Manifest 550* also indicates that inodes 101 and 103 were last modified in version 1 of shard 1. Manifest 550* also turn the update flag on to indicate that modified shard 1* 503-1 contains at least one update.
  • the filer server 710 can determine that inode 102 includes modified data while inodes 101 and
  • file 4 inode
  • the new manifest of shard 1 in local version B includes inodes 101 (unmodified file 1 504- 1), 102 (unmodified file 2 504-2), 103 (unmodified file 3 504-3), and 104 (new file 4).
  • the new manifest indicates that inodes 101-103 were each created in version 1 of shard 1 while inode
  • the new manifest also includes a flag in the "on" state to indicate that version 2 of shard 1 contains at least one update.
  • the filer server 720 can determine that inode 104 is new in version 2 of shard 1 while inodes 101-103 are not new.
  • Figure 8 is a flow chart 80 of a method for sending updated data to cloud storage according to an embodiment.
  • filer server 710 determines which directories or sub-directories have the updated flag flipped to the "on" state in local version A of the shared versioned file system 730.
  • su b-d irectory 2-1 502 is the only directory or sub-directory in which the updated flag is flipped on.
  • filer server 710 determines which shards within the updated directories/sub-directories identified in step 810 have the updated flag flipped to the on state.
  • shard 1 503-1 is the only shard in sub-directory 2-1 503-1 in which the updated flag is on. Since filer server 710 has at least one updated shard, the flow chart 80 proceeds to step 830. I n the circumstance when there are no updated shards, the filer server would return to step 810. The filer server can wait for a short time period (e.g., 30 seconds to 1 minute) before returning to step 810.
  • a short time period e.g. 30 seconds to 1 minute
  • filer server 710 sends a request to operations server 700 for a global lock on shard 1 503-1. If a global lock is available and not in use by another interface or filer server, operations server 700 returns the global lock to Filer A 710. If the global lock is not available, operations server 700 returns a message to the filer server to indicate that the global lock is unavailable. I n that case, the filer server 710 can request a global lock for another updated shard and request the global lock on shard 1 503-1 later. Alternatively, the filer server 710 can continue to request the global lock on shard 1 503-1 u ntil the operations server 700 is able to provide it.
  • the operations server 700 can provide the global lock by sending a message to Filer A that indicates that the global lock request is granted.
  • the operations server 700 also updates a global lock table, a database, or a memory location to indicate that the global lock for shard 1 503-1 is not available and, optionally, to indicate which filer server has the global lock.
  • the operations server 700 can query the global lock table/database/memory location for a given shard to determine if a global lock for that shard is available prior granting a global lock request from a filer server. [00 ] After filer server 710 receives the global lock, the flow chart 80 proceeds to step
  • the filer server 710 identifies the portions of shard 1 503-1 that have updated information. This can be a query for the state of each shard directory entry in the cache of filer server 710 as described below.
  • the available states are cache entry dirty (i.e., the shard directory entry contains updated information since the last shard version), cache entry clean (i.e., the shard directory entry does not contain updated information since the last shard version), or cache entry created (i.e., the shard directory entry did not exist in the last shard version; it was created in the present shard version).
  • the shard directory entries of dirty and created contain new information and need to be sent to the cloud/data store.
  • the filer server determines the portions of the directory entry (e.g., a chunk and/or a manifest of a file) that have been updated.
  • the filer server 710 determines from the updated flags of files 1-3 (504-1 to 504-3) that file 2 504-2 is updated while file 1 504-1 and file 3 504-3 have not been updated.
  • the filer server 710 evaluates the manifest 550 of file 2 version 2 and determines the file version 2 includes chunk 1* 505-1 and pointers to chunk 2 505-2 and chunk 3 505-3. Based on this information, the filer server 710 determines that chunk 1* 505-1 is new/updated and chunks 2 505-2 and 3 505-3 are not new.
  • Data is stored in cloud storage 750 by inode number and version number.
  • the directory inodes/io/Si/vi includes pointers to the contents of shard 1 (i.e., inode 101 (file 1 504-1), inode 102 (file 2 504-2), and inode 1** (file n 504-n)).
  • the pointer to each inode (file) is to the latest version of the inode (file).
  • inode 102 (file 2 504-2) includes a pointer to inodes/102/now.
  • the most recent version of file 2 includes a manifest 510 that identifies inode 1001 (chunk 1 505-1), inode 1002 (chunk 2 505-2), and inode 10** (chu nk n 505 ⁇ ) and the relationship between the chunks (e.g., offset) as the components that form file 2.
  • a manifest 510 that identifies inode 1001 (chunk 1 505-1), inode 1002 (chunk 2 505-2), and inode 10** (chu nk n 505 ⁇ ) and the relationship between the chunks (e.g., offset) as the components that form file 2.
  • filer server 710 sends the update portions of updated shard 1 to the cloud/data store.
  • Filer server 710 can place a local lock on shard 1 during this step.
  • filer server 710 creates a new version (version 2) on cloud storage for shard 1 503-1 at inodes/io/Si/v2.
  • Version 2 of shard 1 includes a new manifest that identifies that the shard includes inodes 101-103 (corresponding to files 1-3). Since no files have been added or deleted from shard 1, the inodes identified in the manifest are the same in versions 1 and 2 of shard 1.
  • the metadata for inode 102 indicates that inode 102 was created in version 1 of shard 1 and last updated in version 2 of shard 1.
  • the metadata for inodes 101 and 103 indicate that they were created in version 1 of shard 1 but have not been updated.
  • filer server 710 creates a new version (version 2) at inodes/102/2.
  • the most recent version of file 2 includes a new manifest 550 that identifies modified inode 1001 (chunk 1* 505-1) and pointers to unmodified inode 1002 (chu nk 2 505-2) and unmodified inode 1003 (chunk 3 505-3) and the relationship between the chunks (e.g., offset) as the components that form version 2 of file 2 504-2.
  • filer server 710 sends modified inode 1001 (chunk 1* 505-1) to the cloud/data store.
  • filer server 710 releases the global lock 860 on shard 1 503-1 back to operations server 700.
  • the release of the global lock 860 can occur by filer server 710 sending a message to operations server 700 to indicate that global lock is released.
  • Filer server 710 can also update an internal memory or cache to indicate (e.g., via a flag or bit) that the filer server 710 no longer has the global lock on shard 1 503-1. Filer server 710 also releases the local lock on shard 1 503-1 if such a lock was placed on shard 1 503-1.
  • the release of the lock lock can occur by updating an internal memory or cache to indicate e.g., via a flag or bit) that the filer server 710 no longer has a local lock on shard 1 503-1.
  • the filer server 710 determines if there are any additional updated shards that need to be sent to the cloud/data store. If so, the flow chart 80 returns to step 830 where the filer server 710 requests a global lock on the next updated shard. If there are no additional updated shards to send to the cloud/data store, the flow chart 80 returns to step 810 to re-start the cloud update process.
  • the filer server 710 can wait for a predetermined time period (e.g., 1 to 5 minutes) before restarting the flow chart 80.
  • operations server 700 maintains a table 90 of such updates as illustrated in Figure 9.
  • Table 90 includes the updated inode and the u pdated shard within the u pdated inode for each update.
  • Table 90 also includes an event number that operations server 700 assigns to each update.
  • Table 90 illustrates that the event number increases by one integer value for each update, though the event number can increase by a different amount in some embodiments. For example, the event number can increase by multiple integers, a decimal (e.g., 100.1, 100.2, etc.), or other unit.
  • the update to shard 1 of inode 10 (sub-directory 2-1 502) described above is included as event number 102 in table 90.
  • Filer servers 710, 720 query the operations server 700 periodically to determine whether there are any recent updates to the shared versioned file system as indicated by the event nu mber. For example, filer server 720 last synchronized updates to the shared versioned file system at event number 100 as illustrated in Figure 8. Since that time, there have been 5 updates to the shared versioned file system, as represented by event numbers 101-105. In order for filer server 720 to update its local version 740 of the global file system with the latest changes, filer server 720 retrieves and merges the updates represented by event numbers 101- 105 into its local version 740 of the global file system.
  • FIG. 10 is a flow chart 1000 of a method for synchronizing updates from a local version of a shared versioned file system to a cloud data store according to an embodiment.
  • the operations server receives a request for a global lock on a shard, such as shard 1 of inode 10.
  • the operations server determines if the global lock is available for the requested shard. If the global lock is available and not in use by another filer server, the operations server sends the global lock to the requesting filer server in step 1030. If the global lock is not available, the operations server can continue to check if for the global lock in an available state. In addition or in the alternative, the operations server can respond to the filer server that the global lock is not available. The filer server can optionally repeat the request for the global lock on the requested shard.
  • the operations server After sending the global lock to the requesting filer server in step 1030, the operations server adds a new event to the update table in step 1040.
  • the update table can be the same or substantially the same as the table illustrated in Figure 9. In general, the update table is a list of each update to a shard in the cloud-based data store. Each update is assigned an event number.
  • the update table can be used by the filer servers to synchronize updates from the cloud-based data store to their respective local versions of the shared versioned operating system. After the requesting filer server has pushed the update directory entries of the requested shard to the cloud-based data store, the operations server receives 1050 the global lock back from the requesting filer server.
  • FIG 11 is a flow chart 1100 of a method for synchronizing updates from a cloud-based a shared versioned file system to a local version of same according to an embodiment.
  • the filer server queries the operating sever for a list of updates to the global file system that have occurred since the last event number updated to the file server.
  • Filer 720 queries the operating server for a list of u pdates that have occurred since event 100, the last event number updated to file server 720.
  • the file server can include the last event number updated to the file server in which case the operating server determines if the file server has the most recent updates by comparing the last event number updated to the file server with the most recent event number on the operations server.
  • the file server server can request the operations server for the most recent event number and the file server r can compare the last event number updated to the file server with the most recent event number on the operations server.
  • step 1120 the file server or operations server determines if there are any new
  • step 1010 If the query in step 1010 includes the last event number updated to the file server, the operations server compares the last event number and the most recent event number to determine if there are any new events.
  • the file server determines if there are any updates by comparing the most recent event on the operations server with the last event number updated to the file server, as discussed above. If there are new events, the file server requests the operations server to provide the inode number and shard number associated with each new event number.
  • step 1120 If the result of step 1120 is that there are no new events since the last event number, the flow chart 1100 returns to step 1110. I n some embodiments, the file server briefly pauses (e.g., for 30 seconds to 1 minute) before returning to step 1110.
  • step 1120 If the result of step 1120 is that there are new events since the last event nu mber, the flow chart 80 proceeds to step 1130.
  • step 1130 the file server receives, for each new event, the inode number and shard number associated with the new event.
  • the new event includes inode 10 (sub-directory 2-1 502) and shard 1 (e.g., in the form of /inode/io/si).
  • each shard includes a manifest of its shard directory entries (e.g., inodes corresponding to files) and metadata about each shard directory entry, such as the version of the shard in which it a file (inode) was created and the version of the shard in which the file (inode) was last updated.
  • the file server uses this metadata in steps 1150 and 1160 to determine the state of each directory entry in the latest cloud version of the shard (step 1150) and the state of each directory entry in the cache version of the shard (step 1160).
  • step 1170 the file server performs the appropriate operation on each cache directory entry according to the table below.
  • the file server determines if there are any additional updated shards received from the operations server that have not been processed. If so, the file server returns to step 1150 to determine the state of each directory entry in the next unprocessed shard. This loop continues until all updated shards received from the operations server have been processed.
  • the filer server in step 1180 returns to step 1110 to query the operation server for updates since the last event number. In this case, the last event event number updated to the filer server would be the last event number from step 1130 in the last iteration through flow chart 1100.
  • the state of a given entry in a cloud shard version can be determined as follows.
  • cache shard 1 includes at least one modified directory entry that needs to be pushed to the cloud, at which point a new event nu mber will be created at the operations server. As a shorthand, this state is referred to as "cache entry dirty.”
  • the filer server performs different operations depending on the state of a directory entry (e.g., File 1) in the cloud shard and in the cache shard. These operations are summarized in Table 1 in Figure 15. The description below continues to use File 1 and shard 1 as a representative directory entry and shard for discussion purposes.
  • a directory entry e.g., File 1
  • Table 1 in Figure 15.
  • the filer server creates a copy of File 1 in a new version of cache shard 1.
  • the filer server determines that there is a conflict. This scenario could occur if users associated with different filer server create a file with the same name in the same directory (shard). I n a conflict state, the filer server saves conflicted version of File 1 from cache shard 1 to the cloud and changes its file name to indicate that it is a conflicted file, as described above.
  • filer server merges the updates from the cloud version of File 1 into the cache version of File 1, as d iscussed herein.
  • This scenario could occur if a user associated with filer server A makes an update to File 1 and sends that update to the cloud while filer server B has a clean copy in cache of the prior version of File 1.
  • filer server B has an old version of File 1 and needs to synchronize with the cloud to obtain the updates to File 1.
  • the filer server determines that there is a conflict. This scenario could occur if a user associated with filer server A makes an u pdate to File i, which already exists in the cloud while a user associated with filer server B deletes File 1 and then creates a new File 1.
  • the filer server saves conflicted cache version of File 1 in shard 1 to the cloud and changes its file name to indicate that it is a conflicted file, as described above.
  • the filer server keeps the cache version of File 1. This scenario could occur if a user creates a file that does not yet exist in the cloud. File 1 will be sent to the cloud the next time the filer server pushes its updates/snapshot to the cloud.
  • Figure 12 is a simplified illustration of a representative shard according to an embodiment.
  • the representative shard in Figure 12 is shard 1 (i.e., Si) of inode 1, which is illustrated in the format of /inodes/[inode number]/[shard number]/[shard version number].
  • version 1 of shard 1 in inode 1 is represented as /inodes/i/Si/i.
  • Figure 12 illustrates the manifest 1200 of shard 1 version 1, which is written in XML (though other hierarchical coding languages can be used).
  • the manifest identifies its inode and shard number using respective ⁇ inode> and ⁇ shard> tags.
  • the manifest also includes a list of directory entries in shard 1 version 1. I n this example, the only d irectory entry is for inode 100, which has the name of filei.txt.
  • the manifest also indicates that inode 100 has a size of 1,024 bytes.
  • Figure 13 is a simplified illustration of the representative shard from Figure 12 after an update according to an embodiment.
  • manifest 1300 of shard 1 version 2 it is apparent that inode 101 (file 2.txt) has been added to shard 1. I node 101 has a size of 2,048 bytes.
  • manifest 1300 includes the directory entries of inode 100 (filei.txt) and inode 101 (file2.txt).
  • Figure 14 is a simplified illustration of a representative d irectory entry in the representative shard of Figure 13.
  • the representative directory entry in Figure 14 is inode 101, which corresponds to file2.txt as discussed above.
  • the directory entry is illustrated in the format of /inodes/[inode number/[inode version number]. Using this format, version 1 of inode 101 is represented as /inodes/ioi/i.
  • the latest version number of a shard or inode can be located in cloud storage by the version number "now.”
  • the manifest 1400 indicates that inode 101 is formed of chunks having a handle (or name) of ci and C2.
  • the manifest 1400 also includes metadata on the relationship between the chunks. I n this case, manifest 1400 indicates that chunk ci has an offset of o and a length of 1,024 bytes.
  • Manifest 1400 also indicates that chunk C2 has an offset of 1,024 and a length of 1024 bytes. In other words, inode 101 has a total length of 2,048 bytes where chunk ci precedes chu nk C2.
  • Chunks ci and C2 each refer to an object in the cloud object store.
  • chunk ci refers to the directory /chunks/ci/data which includes a pointer to the latest version of chunk ci, which in this case is version 1.
  • version 1 of chunk 1 can be found at
  • chunk C2 refers to the d irectory /chunks/c2/data which includes a pointer to the latest version of chunk C2, which in this case is version 1.
  • version 1 of chunk 2 can be found at /chunks/c2/refs/ioo/i.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

Selon la présente invention, un serveur d'opération synchronise les mises à jour sur un système de partage de fichiers par versions en nuage. Le système de partage de fichiers par versions comprend des répertoires et des sous-répertoires qui sont divisés en segments. Le serveur d'opération coordonne des demandes provenant de serveurs de fichiers locaux, chaque serveur exécutant une version locale respective du système de partage de fichiers par versions, en vue de mettre à jour un segment dans le système de partage de fichiers par versions en nuage. Le serveur d'opération peut fournir un verrou global sur le segment à un serveur de fichiers local avant de mettre à jour le segment dans le système de partage de fichiers par versions en nuage.
PCT/US2017/038661 2016-06-22 2017-06-22 Synchronisation de niveau de segment d'une mémoire de données en nuage et de systèmes de fichiers locaux WO2017223265A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US201662353322P 2016-06-22 2016-06-22
US201662353320P 2016-06-22 2016-06-22
US62/353,320 2016-06-22
US62/353,322 2016-06-22

Publications (1)

Publication Number Publication Date
WO2017223265A1 true WO2017223265A1 (fr) 2017-12-28

Family

ID=60784633

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2017/038661 WO2017223265A1 (fr) 2016-06-22 2017-06-22 Synchronisation de niveau de segment d'une mémoire de données en nuage et de systèmes de fichiers locaux

Country Status (1)

Country Link
WO (1) WO2017223265A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110018999A (zh) * 2019-04-15 2019-07-16 深信服科技股份有限公司 一种文件管理方法、系统及电子设备和存储介质
CN110245122A (zh) * 2019-05-08 2019-09-17 华为技术有限公司 一种数据处理的方法和kv存储系统
EP3572923A1 (fr) * 2018-05-23 2019-11-27 Fujitsu Limited Système de traitement d'informations, appareil de traitement d'informations et programme informatique
CN112579698A (zh) * 2020-12-02 2021-03-30 京东数字科技控股股份有限公司 数据同步方法、装置、网关设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080059656A1 (en) * 2006-08-31 2008-03-06 Saliba Bassam A Content synchronization among associated computing devices
US20100191774A1 (en) * 2009-01-23 2010-07-29 Nasuni Corporation Method and system for versioned file system using structured data representations
US20150067109A1 (en) * 2013-08-28 2015-03-05 Beijing Founder Electronics Co., Ltd. Method for processing shared file and cloud storage server
US20150370827A1 (en) * 2014-06-24 2015-12-24 Panzura, Inc. Synchronizing file updates between two cloud controllers of a distributed filesystem
WO2016036288A1 (fr) * 2014-09-02 2016-03-10 Telefonaktiebolaget L M Ericsson (Publ) Procédé, moyen de fonction de synchronisation de nuage et système de fichiers permettant de gérer des fichiers et des répertoires situés dans un service de stockage en nuage

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080059656A1 (en) * 2006-08-31 2008-03-06 Saliba Bassam A Content synchronization among associated computing devices
US20100191774A1 (en) * 2009-01-23 2010-07-29 Nasuni Corporation Method and system for versioned file system using structured data representations
US20150067109A1 (en) * 2013-08-28 2015-03-05 Beijing Founder Electronics Co., Ltd. Method for processing shared file and cloud storage server
US20150370827A1 (en) * 2014-06-24 2015-12-24 Panzura, Inc. Synchronizing file updates between two cloud controllers of a distributed filesystem
WO2016036288A1 (fr) * 2014-09-02 2016-03-10 Telefonaktiebolaget L M Ericsson (Publ) Procédé, moyen de fonction de synchronisation de nuage et système de fichiers permettant de gérer des fichiers et des répertoires situés dans un service de stockage en nuage

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3572923A1 (fr) * 2018-05-23 2019-11-27 Fujitsu Limited Système de traitement d'informations, appareil de traitement d'informations et programme informatique
CN110018999A (zh) * 2019-04-15 2019-07-16 深信服科技股份有限公司 一种文件管理方法、系统及电子设备和存储介质
CN110018999B (zh) * 2019-04-15 2023-07-11 深信服科技股份有限公司 一种文件管理方法、系统及电子设备和存储介质
CN110245122A (zh) * 2019-05-08 2019-09-17 华为技术有限公司 一种数据处理的方法和kv存储系统
CN110245122B (zh) * 2019-05-08 2022-08-09 华为技术有限公司 一种数据处理的方法和kv存储系统
CN112579698A (zh) * 2020-12-02 2021-03-30 京东数字科技控股股份有限公司 数据同步方法、装置、网关设备及存储介质

Similar Documents

Publication Publication Date Title
US11436201B2 (en) Network accessible file server
US11567903B2 (en) Versioned file system with global lock
US9235596B2 (en) Versioned file system with sharing
US11640374B2 (en) Shard-level synchronization of cloud-based data store and local file systems
US9274896B2 (en) Versioned file system with fast restore
US11442902B2 (en) Shard-level synchronization of cloud-based data store and local file system with dynamic sharding
US8990272B2 (en) Versioned file system with pruning
US11726967B2 (en) Systems and methods for restoring an interface to a global file system
US11693819B2 (en) Cloud-native global file system with multi-site support using push classes
WO2017223265A1 (fr) Synchronisation de niveau de segment d'une mémoire de données en nuage et de systèmes de fichiers locaux
WO2017197012A1 (fr) Gestion des versions à l'aide d'un numéro de référence d'événement dans un magasin de données en nuage et des systèmes de fichiers locaux
WO2017196974A1 (fr) Serveur de fichiers accessible par réseau
US11656946B2 (en) Cloud-native global file system with reshapable caching

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17816183

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 01.04.2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17816183

Country of ref document: EP

Kind code of ref document: A1