JP2015194973A - Distributed file system, backup file acquisition method, control device, and management device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a distributed file system contributing to performance improvement during backup processing.SOLUTION: The distributed file system includes a control device for controlling access to a plurality of nodes and a management device for managing information required for backup. The control device transfers a file to the management device when storing the file in a node. The management device includes: a first storage unit in which metadata of the transferred file is stored; a second storage unit in which actual data meeting a prescribed condition is stored; and a file access unit which, in response to reception of a transmission request for metadata from a backup device, acquires the metadata stored in the first storage unit to transmit it to the backup device, and in response to reception of a transmission request for actual data of a file stored in a node, transmits one of actual data stored in the second storage unit and actual data stored in a node in which the file for which the transmission request is received from the backup device is stored, to the backup device.

Description

  The present invention relates to a distributed file system, a backup file acquisition method, a control device, and a management device.

  In recent years, with the increase in the amount of data handled by information processing devices such as servers, the importance of a distributed file system that distributes data to a plurality of nodes connected to a network and stores the data at each node is increasing. . In a distributed file system or a distributed NAS (Network Attached Storage), for example, as shown in FIG. 13, a large number of NAS and NAS-like network storages are distributed on the network as distributed nodes 62-1 to 62-3. .

  In the distributed file system shown in FIG. 13, the file system integration device 61 processes an access request from the client 60 to the distributed nodes 62-1 to 62-3. The file system integration device 61 includes a file access unit 300 that serves as an interface for the client 60 and the backup device 63, and a distributed storage unit 301 that distributes and stores files in the respective distributed nodes 62-1 to 62-3. In the distributed file system, by distributing and arranging nodes that store files, it is possible to execute processes related to storage of files (data) in parallel, and by providing redundancy as necessary, The overall system capacity, performance and reliability are improved.

  Here, not only in a distributed file system, but in a system that stores about a file, backup of the stored file is important. However, with regard to file backup in a distributed file system, the amount of data handled may be enormous, and a typical backup method has not been established.

  Patent Document 1 discloses a system in which an index server that centrally manages files in a plurality of file servers is installed, and the index server determines a file duplication destination and a movement destination.

  Patent Document 2 discloses a file system that duplicates (replicates) the entire distributed file system and limits the files to be copied to a part.

JP 2010-218335 A JP 2004-252957 A

  Each disclosure of the above prior art document is incorporated herein by reference. The following analysis was made by the present inventors.

  A typical file stem backup process is performed as follows. First, the entire file system is scanned to create a list of files to be backed up. At that time, the metadata is often referred to in order to confirm the path name and update time of the file. Note that the metadata is file management data, and includes information (attribute values) relating to the file update date, access authority, size, and the like. Second, the actual data of the backup target file is acquired from the file system. The actual data is the contents of the file itself. Third, the acquired actual data is stored in the storage device.

  When the above backup processing is executed in a distributed file system, the following problems may occur.

  In the distributed file system, the file is distributed and stored in each distributed node. Therefore, for each distributed unit, the process of specifying the distributed node storing data and the process of switching the distributed node to be read are performed. Necessary. Such processing causes overhead such as starting connection with each distributed node and ending connection. That is, an overhead peculiar to the distributed file system occurs even when reading data of either metadata or actual data.

  In addition, even if the total amount of data is the same, when accessing actual data, the smaller the individual file size, the greater the overhead. This is because, in order to read a large number of files, switching of distributed nodes and more processing for acquiring data occur.

  In order to solve the above problem, for example, it is conceivable to use a technique (centralized data management) as disclosed in Patent Document 1. Specifically, as shown in FIG. 14, a metadata storage device 65 accessible from the metadata management unit 402 of the file system integration device 64 is provided, and the metadata is centrally managed (centralized management) in the storage device. It is possible. Such a technique is an effective technique for determining a snapshot that is an instantaneous image of the entire file system and defining a quiesce point, and is mainly used in a NAS virtualization apparatus or the like.

  However, centralized management of metadata can be a bottleneck in performance as well as a single point of failure (SPF) in the distributed file system. Further, the above method does not contribute to improvement of access performance for real data.

  As described above, in the distributed file system, the occurrence of overhead during backup processing caused by distributing and storing files, that is, the performance degradation of the distributed file system becomes a problem.

  It is an object of the present invention to provide a distributed file system, a backup file acquisition method, a control device, and a management device that contribute to performance improvement during backup processing.

  According to a first aspect of the present invention, comprising: a control device that controls access to a plurality of nodes that store files; and a management device that manages information required when processing a request from a backup device, The control device includes a transfer unit that transfers the file to be stored to the management device when the file is stored in at least one of the plurality of nodes, and the management device receives the transfer A first storage unit that stores file metadata, a second storage unit that stores actual data satisfying a predetermined condition among the actual data of the transferred file, and the node from the backup device When a metadata transmission request regarding a file to be stored is received, the metadata stored in the first storage unit is acquired and stored in the backup device. And when receiving a request for transmission of actual data of the file stored in the node from the backup device, the actual data stored in the second storage unit and the file received from the backup device are transmitted. A distributed file system is provided, comprising: a file access unit that transmits any of the actual data stored in the storing node to the backup device.

According to a second aspect of the present invention, the control apparatus includes a control device that controls access to a plurality of nodes that store files, and a management device that manages information required when processing a request from a backup device. A method for acquiring a backup file of a distributed file system, the step of transferring the file to be stored to the management device when storing the file in at least one of the plurality of nodes; Storing metadata of the transferred file in a first storage unit; storing actual data satisfying a predetermined condition in the second storage unit among the actual data of the transferred file; When receiving a metadata transmission request regarding the file stored in the node from the backup device, the first storage unit stores the metadata. The step of acquiring metadata and transmitting it to the backup device, and the actual data stored in the second storage unit when receiving a request for transmitting the actual data of the file stored in the node from the backup device And a step of transmitting to the backup device any of actual data stored in a node that stores a file that has received a transmission request from the backup device.
Note that this method is linked to specific machines such as a control device and a management device.

  According to the third aspect of the present invention, the file to be stored is controlled when the access to the plurality of nodes storing the file is controlled and the file is stored in at least one of the plurality of nodes. Is transferred to a management device that manages information required for processing a request from the backup device.

  According to the fourth aspect of the present invention, the management device manages information required when processing a request from a backup device, and is transferred from a control device that controls access to a plurality of nodes storing files. From the first storage unit that stores the metadata of the file, the second storage unit that stores the actual data satisfying a predetermined condition among the actual data of the file transferred from the control device, and the backup device, When receiving a metadata transmission request regarding a file stored in the node, the metadata stored in the first storage unit is acquired and transmitted to the backup device, and the node stores the metadata from the backup device. When a request to send actual data of a file to be received is received, the actual data stored in the second storage unit and the backup data are sent from the backup device. Any node storing the actual data for storing requests received file is transmitted to the backup device, comprising: a file access unit, a management device is provided.

  According to each aspect of the present invention, a distributed file system, a backup file acquisition method, a control device, and a management device that contribute to performance improvement during backup processing are provided.

It is a figure for demonstrating the outline | summary of one Embodiment. It is a figure which shows an example of a structure of the distributed file system which concerns on 1st Embodiment. It is a sequence diagram which shows an example of an operation | movement when there exists a file write request from a client. It is a sequence diagram which shows an example of the operation | movement which concerns on the snapshot production | generation of a distributed file system. It is a flowchart which shows an example of the operation | movement which concerns on the snapshot acquisition process of a file system integration apparatus. It is a sequence diagram which shows an example of the operation | movement which concerns on determination of the backup file of a distributed file system. It is a sequence diagram which shows an example of the operation | movement which concerns on the backup of the actual data of a distributed file system. It is a flowchart which shows an example of the operation | movement at the time of processing the backup file transmission request of a replication file system management apparatus. It is a figure which shows an example of a structure of a distributed file system. It is a figure which shows an example of a structure of a distributed file system. It is a figure for demonstrating an example of a backup system. It is a figure which shows an example of a structure of the distributed file system which concerns on 2nd Embodiment. It is a figure which shows an example of a structure of a distributed file system. It is a figure which shows an example of a structure of a distributed file system.

  First, an outline of an embodiment will be described with reference to FIG. Note that the reference numerals of the drawings attached to the outline are attached to the respective elements for convenience as an example for facilitating understanding, and the description of the outline is not intended to be any limitation.

  As described above, a distributed file system that contributes to performance improvement during backup processing is desired.

  Therefore, the distributed file system shown in FIG. 1 is provided as an example. The distributed file system includes a control device 100 that controls access to a plurality of nodes that store files, and a management device 101 that manages information required when processing a request from a backup device. The control device 100 includes a transfer unit 102 that transfers a file to be stored to a management device when storing the file in at least one of a plurality of nodes. The management apparatus 101 includes a first storage unit 103 that stores metadata of the transferred file, and a second storage unit 104 that stores actual data that satisfies a predetermined condition among the actual data of the transferred file. When a transmission request for metadata relating to a file stored in the node is received from the backup device, the metadata stored in the first storage unit 103 is acquired and transmitted to the backup device. When a transmission request for the actual data of the file to be stored is received, either the actual data stored in the second storage unit 104 or the actual data stored in the node storing the file that has received the transmission request from the backup device is stored. And a file access unit 105 for transmitting to the backup device.

  The management apparatus 101 uses a file transferred from the control apparatus 100 to construct a copy of a distributed file system composed of a plurality of distributed nodes, and to create various requests from the backup apparatus. Process using file system replication. The management apparatus 101 basically stores metadata of all files when constructing a replica of the file system, but the actual data satisfying a predetermined condition (for example, the file size is equal to or smaller than a threshold). Store the data. Therefore, the management apparatus 101, when receiving a metadata acquisition request or a transmission request of actual data stored therein from the backup apparatus, copies the file system without accessing the main body of the distributed file system including the distributed nodes. It can be dealt with by accessing. As a result, overhead when performing backup of the distributed file system is reduced, and performance improvement during backup processing can be achieved.

  Hereinafter, specific embodiments will be described in more detail with reference to the drawings.

[First Embodiment]
The first embodiment will be described in more detail with reference to the drawings.

  FIG. 2 is a diagram illustrating an example of the configuration of the distributed file system according to the first embodiment. The distributed file system includes a client 1, a file system integration device 2, a plurality of distributed nodes 3-1 to 3 -n (n is a positive integer, hereinafter the same), a backup device 4, and a duplicate file system management device 5. And comprising. Each device included in the distributed file system shown in FIG. 2 is connected to each other via a network. In the following description, when there is no particular reason for distinguishing the distributed nodes 3-1 to 3-n, they are simply expressed as “distributed node 3”.

  The client 1 is a computer or the like used by a user who uses the distributed file system for business or the like. The client 1 makes a file write request or read request to the file system integration device 2.

  The file system integration device 2 is a device that processes an access request from the client 1. Specifically, the file system integration device 2 is a device that controls access to the distributed nodes 3 that store files. The file system integration device 2 corresponds to the control device 100 described above.

  When the file system integration device 2 makes a file write or update request to the distributed node 3, the file system integration device 2 also makes an equivalent request to the replication file system management device 5. Further, the file system integration device 2 has a function of causing the distributed node 3 and the replication file system management device 5 to acquire a snapshot at an arbitrary timing.

  The distributed node 3 is a device that stores files written from the file system integration device 2. The distributed node 3 has a function of acquiring a snapshot at a specific timing in response to a notification from the file system integration device 2.

  The backup device 4 includes a medium that actually stores backup data (for example, a tape device, an archive device, etc.), and corresponds to a computer in which software for managing and executing backup of a distributed file system is installed.

  The duplicate file system management device 5 is a device that processes various requests from the backup device 4. The duplicate file system management device 5 relates to a file metadata copy stored in the entire distributed file system and a file satisfying a predetermined condition among the files stored in the distributed node 3 in order to respond to the request from the backup device 4. At least a copy of the actual data is stored. The duplicate file system management apparatus 5 corresponds to the management apparatus 101 described above.

  Referring to FIG. 2, the file system integration device 2 includes a file access unit 10, a distributed storage unit 11, a file transfer unit 12, and a snapshot control unit 13. Each unit such as the file access unit 10 can also be realized by a computer program that causes a computer mounted in the file system integration apparatus 2 to execute processing that will be described in detail later using the hardware thereof.

  The file access unit 10 is a means for providing an interface with the client 1. In accordance with the access request from the client 1, the file access unit 10 delivers the file acquired from the client 1 to the distributed storage unit 11 and the file transfer unit 12, and issues a file write request and a file update request.

  The distributed storage unit 11 is means for storing a file in the distributed node 3 in response to a file write request from the file access unit 10. Specifically, the distributed storage unit 11 transmits a write request for file data or each fragment thereof to the distributed node 3. The distributed storage unit 11 determines a distributed node 3 that transmits a write request based on a distributed storage rule of a predetermined distributed file system.

  When the file system integration device 2 reads a file stored in the distributed node 3, a reading unit (not shown) is used.

  When the distributed storage unit 11 stores a file in at least one or more distributed nodes 3 among the distributed nodes 3-1 to 3 -n, the file transfer unit 12 stores the file to be stored in a duplicate file system management apparatus. 5 is a means for transferring the data. That is, the file transfer unit 12 issues a file write request equivalent to that performed by the distributed storage unit 11 to the distributed node 3 to the duplicate file system management apparatus 5 and stores (stores) the file. A node number for specifying the distributed node 3 to be transmitted is transmitted to the replication file system management apparatus 5.

  The snapshot control unit 13 is means for causing the distributed node 3 and the replication file system management apparatus 5 to acquire a snapshot at an arbitrary timing. The snapshot control unit 13 has a function of managing file generations at the time when a snapshot acquisition instruction (to be described later) is received.

  The distributed node 3 includes a data storage unit 20 and a metadata storage unit 21. The data storage unit 20 is means for storing the actual data of the file. The metadata storage unit 21 is a means for storing file metadata.

  The duplicate file system management apparatus 5 includes a file access unit 30, a snapshot unit 31, and a duplicate file system unit 32. Each unit such as the file access unit 30 can also be realized by a computer program that causes a computer mounted in the duplicate file system management apparatus 5 to execute processing that will be described in detail later using the hardware thereof.

  Both the file access unit 30 and the snapshot unit 31 function as an interface to the backup device 4. Specifically, the file access unit 30 is means for processing a data transmission request from the backup device 4.

  The snapshot unit 31 is means for processing a request relating to snapshot generation from the backup device 4.

  The duplicate file system unit 32 specifies a copy of the metadata of all files stored in the distributed file system, a copy of actual data regarding some files stored in the distributed file system, and a distributed node that stores the remaining actual data. It is a means to manage information to do. In other words, the duplicate file system management apparatus 5 constructs and manages a virtual file system for the file system main body by having a copy of the distributed file system main body composed of the distributed nodes 3 therein.

  It can be considered that the duplicate file system management apparatus 5 stores the actual data of all files targeted by the distributed file system. However, if the replicated file system management device 5 tries to store all data stored in the entire distributed file system, the same capacity (storage medium having the same capacity) as the capacity of the distributed file system itself is required. Therefore, in the duplicate file system management apparatus 5 according to the first embodiment, a file that satisfies a predetermined condition is stored in the duplicate file system management apparatus 5, and a file that does not satisfy the condition is stored in the distributed node 3.

  The predetermined condition can be, for example, a file size. More specifically, it is a condition that a large file is stored in the distributed node 3 and a small file is also stored in the duplicate file system management apparatus 5. Actually, a threshold value is set in advance, and whether or not actual data is stored in the duplicate file system management apparatus 5 is determined based on the threshold value. In the present embodiment, a description will be given assuming that a file having a size equal to or smaller than a threshold value is stored (replicated) in the replicated file system management apparatus 5. However, the predetermined condition is not limited to the file size. For example, the file type (for example, text file or binary file) or the like may be set as the predetermined condition.

  The duplicate file system unit 32 includes an indirect data storage unit 201, an indirect data reference unit 202, a transfer actual data storage unit 203, an actual data reference unit 204, and a transfer metadata storage unit 205. .

  The indirect data storage unit 201 stores information (hereinafter referred to as indirect data) necessary for specifying the distributed node 3 in which the file is stored among the plurality of distributed nodes 3 constituting the distributed file system. . The indirect data for specifying the distributed node 3 in which the file is stored is information that is appropriately selected according to the data storage method of the distributed file system. For example, the indirect data storage unit 201 stores information such that a file identifier and a node number of the distributed node 3 storing the file are paired.

  The indirect data reference unit 202 is a means for accessing a file stored in the distributed node 3 based on information stored in the indirect data storage unit 201.

  The transfer actual data storage unit 203 stores the actual data of a file that satisfies a predetermined condition among the files transferred from the file system integration device 2. In the present embodiment, as described above, the file whose size is equal to or smaller than the threshold value is stored in the transfer actual data storage unit 203.

  The actual data reference unit 204 is a means for accessing a file stored in the transfer actual data storage unit 203.

  The transfer metadata storage unit 205 stores metadata of all files transferred from the file system integration device 2. Compared with actual data, the size of the metadata is extremely small, and even if metadata of all data stored in the distributed node 3 is stored, the capacity required for storage is limited.

  Next, the operation of the distributed file system will be described. The description of the operation of the distributed file system will be made for each function, such as writing a file to the distributed node 3 and performing a backup from the distributed node 3.

<Response to file write request from client 1>
FIG. 3 is a sequence diagram illustrating an example of an operation when a file write request is issued from the client 1.

  In step S <b> 101, the client 1 transmits a file write request to the file system integration apparatus 2.

  In step S <b> 102, the distributed storage unit 11 of the file system integration device 2 requests the distributed node 3 to store file data or each fragment thereof in response to the request.

  At that time, the file transfer unit 12 of the file system integration device 2 transfers the file to be stored in the distributed node 3 to the replication file system management device 5 (step S103). That is, the file transfer unit 12 sends a file write request equivalent to that performed by the distributed storage unit 11 to the duplicate file system management apparatus 5.

  In FIG. 3, it is described that step S103 is executed after step S102. However, this is not intended to limit the execution order of the processes, and these execution orders may be reversed or in parallel (simultaneously). May be executed.

  In step S <b> 104, the distributed node 3 stores actual data in the data storage unit 20 and metadata corresponding to the metadata storage unit 21 based on the received write request.

  In step S <b> 105, the duplicate file system unit 32 stores the received file metadata in the transfer metadata storage unit 205 in response to the file write request received from the file transfer unit 12.

  In step S106, the duplicate file system unit 32 determines whether or not the size of the transfer file is larger than the threshold, and based on the determination result, stores the data in the indirect data storage unit 201 or the transfer actual data storage unit 203. One of the actual data is stored (steps S107 and S108).

  When the file size of the transfer file is larger than the threshold, the duplicate file system unit 32 stores information (indirect data) for specifying the distributed node 3 in which the file is stored in the indirect data storage unit 201 ( Step S107).

  On the other hand, when the file size is equal to or smaller than the threshold value, the duplicate file system unit 32 stores the actual data satisfying the predetermined condition in the transfer actual data storage unit 203.

  The operation described with reference to FIG. 3 is in response to a file write request from the client 1. Even when there is a file update request from the client 1, the operation according to FIG. 3 is performed.

  However, in this case, since the target file already exists in the distributed node 3, the operation of the distributed node 3 in step S104 is “file update”. Further, since the metadata of the update file also exists in the duplicate file system management apparatus 5, the operation of the apparatus in step S105 is “metadata update”.

  On the other hand, the size of the update file may change, and it is not certain whether the update file satisfies a predetermined condition. Therefore, the duplicate file system management apparatus 5 executes again the operations according to steps S106 to S108 for the update file.

<Get snapshot>
The backup device 4 performs backup of the distributed file system according to a schedule predetermined by a distributed file system administrator or the like. Prior to performing the backup, the backup device 4 instructs the replication file system management device 5 to generate a snapshot.

  FIG. 4 is a sequence diagram illustrating an example of operations related to snapshot generation of the distributed file system.

  In step S <b> 201, the backup device 4 transmits a snapshot generation instruction to the duplicate file system management device 5.

  In step S <b> 202, the snapshot unit 31 instructs the file system integration apparatus 2 to acquire (collect) a snapshot.

  In step S203, the file system integration device 2 executes processing related to snapshot acquisition.

  FIG. 5 is a flowchart illustrating an example of an operation related to the snapshot acquisition process of the file system integration apparatus 2. With reference to FIG. 5, an operation related to snapshot acquisition by the file system integration apparatus 2 will be described.

  First, the snapshot control unit 13 manages the generation of a file from the point in time when the snapshot acquisition instruction is received (step S301). Specifically, the snapshot control unit 13 recognizes that generation A is completed and generation B is started thereafter when a snapshot acquisition instruction is received.

  After receiving the snapshot acquisition instruction, the snapshot control unit 13 assigns a tag indicating generation B to the file to be transmitted to the distributed node 3 and the replication file system management apparatus 5 (step S302). Note that a generation A tag is assigned to a file transmitted at the time of a file write / file update request prior to reception of the snapshot acquisition instruction.

  Next, the snapshot control unit 13 causes the file transfer unit 12 to interrupt transmission of a write / update request regarding a file with the generation B tag to the replication file system management apparatus 5 (step S303). By interrupting the transmission of the file with the generation B tag, the process waits for the replication file system management apparatus 5 to complete the write processing and update processing for the file with the generation A tag (step S304). ).

  Next, the snapshot control unit 13 instructs the replication file system management apparatus 5 to acquire a snapshot (step S305). Upon receiving the snapshot acquisition instruction, the duplicate file system management apparatus 5 acquires snapshots of the transfer actual data storage unit 203 and the transfer metadata storage unit 205.

  Thereafter, the snapshot control unit 13 receives a response regarding the completion of snapshot acquisition from the replication file system management apparatus 5 (step S306). After receiving the response, the file transfer unit 12 can transmit a write / update request for the file with the generation B tag to the replication file system management apparatus 5.

  The snapshot control unit 13 performs processing related to snapshot acquisition for the distributed node 3 in parallel with the operations of steps S303 to S306 (step S307). That is, the snapshot control unit 13 performs the same processing as steps S303 to S306 on the distributed node 3.

  Specifically, the snapshot control unit 13 causes the distributed storage unit 11 to interrupt transmission of a write / update request regarding a file with a generation B tag to each distributed node 3. The snapshot control unit 13 interrupts the transfer of the file with the generation B tag so that the writing process and the update process for the file with the generation A tag are completed in each distributed node 3. stand by.

  Thereafter, the snapshot control unit 13 instructs the distributed node 3 that has completed processing related to the writing and update request of the file with the generation A tag to acquire a snapshot. The distributed node 3 that has received the snapshot acquisition instruction acquires a snapshot of the data storage unit 20 and the metadata storage unit 21.

  Thereafter, the snapshot control unit 13 receives a response regarding completion of snapshot acquisition from the distributed node 3. In addition, after receiving a response regarding the completion of snapshot acquisition, the distributed storage unit 11 transmits a write / update request for a file with a generation B tag to the distributed node 3 that has transmitted the response.

  In this way, the snapshot control unit 13 confirms that the processing related to the old generation file among the files transferred to the replication file system management apparatus 5 has been completed, and then the replication file system management apparatus 5 Instruct the snapshot acquisition. The snapshot control unit 13 issues a similar snapshot acquisition instruction to the distributed node 3 as well.

  Returning to the description of FIG. 4, the file system integration apparatus 2 that has received the response regarding the completion of snapshot acquisition from the replication file system management apparatus 5 and the distributed node 3 sends a snapshot acquisition completion notification to the replication file system management apparatus 5. Transmit (step S204).

  The duplicate file system management apparatus 5 that has received the notification notifies the backup apparatus 4 that the generation of the snapshot has been completed (step S205).

<Determining backup target>
When the creation of the snapshot ends normally, the backup device 4 determines a file to be backed up.

  FIG. 6 is a sequence diagram illustrating an example of an operation related to determination of a backup file in the distributed file system.

  In step S401, the backup device 4 requests the duplicate file system management device 5 for a list including attribute information of files stored in the distributed file system.

  In step S402, the file access unit 30 of the duplicate file system management apparatus 5 requests the duplicate file system unit 32 to process the information request. The duplicate file system unit 32 acquires attribute information (for example, update date / time of the file) of each file from the transfer metadata storage unit 205 and returns the acquired information to the file access unit 30.

  The file access unit 30 returns the received file attribute information (list information) to the backup device 4. That is, the file access unit 30 acquires the metadata stored in the transfer metadata storage unit 205 of the duplicate file system unit 32 and responds to the backup device 4.

  The backup device 4 determines a file to be backed up based on the acquired file attribute information (step S403). For example, the backup device 4 sets a file that has been updated since the time of the previous backup as a backup target.

<Actual data backup>
Next, the backup device 4 makes a transmission request regarding the actual data of the backup file to the duplicate file system management device 5.

  FIG. 7 is a sequence diagram illustrating an example of operations related to actual data backup in the distributed file system.

  In step S501, the backup device 4 requests the replication file system management device 5 to transmit actual data related to the backup file. Upon receiving the request, the file access unit 30 of the duplicate file system management apparatus 5 requests the duplicate file system unit 32 to process the transmission request (step S502).

  FIG. 8 is a flowchart showing an example of the operation when the replication file system management apparatus 5 processes the backup file transmission request. The operation of processing a backup file transmission request by the duplicate file system management apparatus 5 will be described with reference to FIG.

  In step S601, the duplicate file system unit 32 determines whether the requested backup file satisfies a predetermined condition. Specifically, the duplicate file system unit 32 determines whether the requested file size of the back file is equal to or smaller than a predetermined threshold.

  When the file size of the requested back file is larger than a predetermined threshold (step S601, No branch), the duplicate file system unit 32 operates the indirect data reference unit 202. Specifically, in step S602, the indirect data reference unit 202 identifies the distributed node 3 storing the requested file based on the information stored in the indirect data storage unit 201, and the distributed node 3 On the other hand, a request for transmission of actual file data in the latest snapshot generation is made.

  In step S <b> 603, the indirect data reference unit 202 responds to the file access unit 30 with the actual data acquired from the distributed node 3.

  On the other hand, when the file size of the requested back file is equal to or smaller than a predetermined threshold (step S601, Yes branch), the duplicate file system unit 32 operates the actual data reference unit 204. Specifically, in step S604, the actual data reference unit 204 responds to the file access unit 30 with the actual data of the file in the latest snapshot generation stored in the transfer actual data storage unit 203.

  In step S <b> 605, the file access unit 30 transmits the actual data acquired from the indirect data reference unit 202 or the actual data reference unit 204 to the backup device 4.

  As described above, when the file access unit 30 receives a request for transmission of the actual data of the backup file from the backup device 4, the actual data or the transfer actual data stored in the distributed node 3 based on a predetermined condition. Any real data stored in the storage unit 203 is transmitted to the backup device 4.

  Returning to FIG. 7, in step S503, the backup device 4 backs up the received actual data.

  In the first embodiment, the file system integration apparatus 2 has been described with respect to all file write requests received from the client 1 and transfers the file to the duplicate file system management apparatus 5. However, the present invention is not limited to transferring all files to the duplicate file system management apparatus 5. For example, even if the file is stored in the distributed node 3 and is not required to be backed up, it is not necessary to transfer it to the duplicate file system management apparatus 5.

  Furthermore, in the first embodiment, the transfer metadata storage unit 205 of the duplicate file system management apparatus 5 has been described with respect to the form of storing the metadata of all transferred files. However, the transfer metadata storage unit 205 is not limited to storing the metadata of all files. For example, in the case of a file that is previously determined not to be a backup target, it is not necessary to store the metadata of the transferred file.

  As described above, in the distributed file system according to the first embodiment, metadata and small-sized files are collected in a single node (replicated file system management apparatus 5). Therefore, access to the distributed node 3 is not required when acquiring such information. As a result, overhead when performing backup of the distributed file system is reduced, and performance improvement during backup processing (for example, faster backup) can be achieved.

  Further, although not necessarily limited to a distributed file system, the backup load may affect the performance of a job that uses the distributed file system. As a technique for avoiding such an influence, there is a technique for separating work load and backup load by utilizing system redundancy.

  For example, as shown in FIG. 9, the distributed nodes 3a, 3c, and 3e are duplicated to generate distributed nodes 3b, 3d, and 3f. When the file system integration device 2a performs normal access to the distributed node (access shown by a solid line), the file system integration device 2a accesses the replication source distributed node. On the other hand, the file system integration apparatus 2a accesses the replicated distributed node when performing access (access indicated by a dotted line) for reading backup data from the distributed node.

  Alternatively, as shown in FIG. 10, it can be considered that the same file A is stored in each of the data storage units 20b included in the distributed nodes 3g to 3i to separate normal access and backup access. Specifically, backup access and normal access traffic are separated, and in normal business and backup, another node may dispatch the processing of the file system integration apparatus 2b so that it responds to a data request.

  In this way, by using the redundancy of the system and avoiding the load contention in the distributed nodes, it is possible to mitigate the influence on the business due to the backup load. However, even with such a method, load contention in the file system integration apparatuses 2a and 2b cannot be resolved.

  On the other hand, in the distributed file system according to the first embodiment, the duplicate file system management apparatus 5 includes the independent transfer metadata storage unit 205 and the file access unit 30, so that access at the time of backup is completely separated. However, processing such as listing of files to be backed up and referring to attribute values can be performed using the built-in metadata. That is, since it is possible to respond to the request of the backup device 4 independently of the distributed file system main body composed of the file system integration device 2 and the distributed node 3, the influence on the business at the time of backup can be reduced.

  Further, in the block level storage which is a similar area, in order to cope with the problem of competition with business, as shown in FIG. 11, the entire data storage unit 51 under the control of the block storage control unit 50 is transferred to another block storage. It is conceivable to replicate in the data storage unit 51a under the control of the control unit 50a. In this case, access from the backup device 4c can be completely separated, and the load of normal work and backup can be separated. However, in such a correspondence, it is necessary to prepare two sets (or more) of the entire device, which is not economical.

  On the other hand, in the distributed file system according to the first embodiment, the entire device is not duplicated, and separation of backup and normal business can be realized.

  Further, as disclosed in Patent Document 2, if the files to be replicated are limited to a part of the files, the capacity and equipment required on the replication side can be suppressed. However, the technique disclosed in Patent Document 2 cannot be applied when the backup target cannot be narrowed down in advance according to a specific standard.

  On the other hand, in the distributed file system according to the first embodiment, the duplicate file system management apparatus 5 narrows down the file copy based on a predetermined condition, and the files that are leaked from the narrow down are obtained from the distributed node 3, so that the backup There is no limitation on the subject.

  Further, the backup data needs to be consistent as a file system, but in a distributed file system composed of a plurality of nodes, it is difficult to back up consistent data as a whole. Further, in order to make backup more meaningful, it is necessary to fix the content of stored data so that it does not change during backup processing. More specifically, the situation where the file A is backed up but the file B stored in the distributed file system before the file A is not backed up must be avoided.

  Therefore, a snapshot that is an image of the entire file system is acquired, and the data of the snapshot is used as a backup target. However, since the distributed file system is composed of a plurality of nodes, a common moment (timing) cannot be defined between the nodes, and it is difficult to obtain a snapshot.

  In the distributed file system according to the first embodiment, since the replicated file system management apparatus 5 employs a centralized NAS configuration, the “snapshot acquisition time on the virtual file system side where the distributed file system is replicated” Since a quiesce point can be defined as, backup can be performed at the quiesce point as a distributed file system.

  In addition, by enabling access from the duplicate file system management apparatus 5 to the distributed node 3, it is not necessary to store actual data relating to all files to be stored in the distributed file system in the duplicate file system management apparatus 5. Therefore, a high-performance, high-reliability, large-capacity NAS for acquiring a copy of the distributed file system in the centralized NAS is not required.

[Second Embodiment]
Next, a second embodiment will be described in detail with reference to the drawings.

  FIG. 12 is a diagram illustrating an example of the configuration of the distributed file system according to the second embodiment. 12, the same components as those in FIG. 2 are denoted by the same reference numerals, and the description thereof is omitted.

  The file system integration device 2c shown in FIG. 12 is different from the file system integration device 2 according to the first embodiment in that an update information storage unit 14 is provided. The duplicate file system management apparatus 5a is different from the duplicate file system management apparatus 5 according to the first embodiment in that it includes a file update history storage unit 206. Further, in the duplicate file system management apparatus 5a, the indirect data storage unit 201 does not exist.

  The update information storage unit 14 shown in FIG. 12 functions as a buffer for files transferred to the duplicate file system management apparatus 5a.

  Here, in response to a file write / update request from the client 1, the file transfer unit 12 transfers the file to the duplicate file system management apparatus 5a, confirms the response, and returns the response to the client 1. The response performance to the client 1 may be deteriorated. Therefore, in the second embodiment, a file buffer to be transferred to the replication file system management apparatus 5a is provided in the file system integration apparatus 2c, and replication in the replication file system management apparatus 5a is executed asynchronously. Yes.

  The buffer size of the update information storage unit 14 is determined according to the performance of the component devices and the assumed access frequency from the client 1, but it is also conceivable to make the buffer size variable. When the buffer of the update information storage unit 14 is full, the writing to the replication file system management device 5a is shifted to the synchronous mode, or the administrator is notified to prompt the countermeasure.

  In the second embodiment, the indirect data storage unit 201 is deleted. This assumes the following two cases.

  In a distributed file system, methods for determining the storage location of each file (or file fragment) can be broadly classified into two types. The first method is a method in which a node for uniquely storing data is determined from attributes such as the file name of each file. The second method is a method in which a node for storing data is not uniquely determined from existing file attributes or the like. In the following description, the first method is referred to as a deterministic method, and the second method is referred to as a non-deterministic method.

  The distributed file system described in the first embodiment assumes a nondeterministic method. Therefore, it is necessary to store node information related to the distributed node 3 that stores individual files in the indirect data storage unit 201.

  When the deterministic method is adopted, the distributed node 3 in which necessary data is stored can be calculated from attribute data stored as normal metadata such as a file name. Therefore, in the case of adopting a definitive method, the indirect data storage unit 201 is not necessary. This is the first case. When the deterministic method is adopted, the indirect data reference unit 202 identifies the distributed node 3 in which the actual data is stored from the attribute data stored in the transfer metadata storage unit 205, and the distributed node Actual data will be obtained from

  On the other hand, even if the non-deterministic method is adopted, if some performance degradation is permitted, the indirect data reference unit 202 performs actual data via the file system integration device 2c in the same manner as the client 1. Is possible to get. By acquiring actual data via the file system integration device 2c, complicated information such as storing information about the distributed node 3 in the indirect data storage unit 201 and maintaining it consistent with the distributed file system main body is unnecessary. It becomes. This is the second case.

  In the second case, the indirect data reference unit 202 requests the file system integration device 2c for actual data of the corresponding file, and acquires data from the file system integration device 2c.

  Here, a part of the functions of the backup device 4 may be realized by the duplicate file system management device 5a. For example, instead of the backup device 4 deciding the backup target, a configuration in which the duplicate file system management device 5a provides the backup device 4 with the substance of the difference file from the previous backup is also conceivable. When performing differential backup in such a form, it is necessary to check all metadata and select a backup target file. That is, in the configuration described with reference to FIG. 2 in the first embodiment, it is necessary to check all the metadata stored in the transfer metadata storage unit 205.

  However, the duplicate file system management device 5a checks the transfer metadata storage unit 205 by storing the update history of the files transferred from the file system integration device 2c in the file update history storage unit 206. In addition, it is possible to efficiently select files that have been updated since the last backup acquisition time and determine the backup target.

  As described above, in the distributed file system according to the second embodiment, by using the update information storage unit 14, the response performance to the client 1 is improved, and the replication file system management device 5a is temporarily stopped. Resilience against response delay can be added. Further, in the duplicate file system management apparatus 5a, by storing the file update history in the file update history storage unit 206, the differential file to be backed up can be easily identified, and the backup performance is further improved.

  The distributed file system described in the first and second embodiments is effective in an environment where backup is efficiently performed while suppressing the influence on normal business (performance degradation, business suspension) as much as possible. It is preferable to operate.

  The configuration and operation of the distributed file system described in the first and second embodiments are merely examples, and various modifications can be made. For example, all or some of the functions of the duplicate file system management apparatus 5 may be incorporated in the file system integration apparatus 2. As an example, if the determination process of step S106 shown in FIG. 3 is incorporated in the file transfer unit 12 and a predetermined condition is not satisfied, transfer of the corresponding file from the file system integration apparatus 2 to the duplicate file system management apparatus 5 is omitted. It is possible to do.

  A part or all of the above embodiments can be described as in the following supplementary notes, but is not limited thereto.

[Appendix 1]
A control device for controlling access to a plurality of nodes storing files;
A management device that manages information required to process a request from the backup device;
Including
The controller is
A transfer unit that transfers the file to be stored to the management device when storing the file in at least one of the plurality of nodes;
The management device
A first storage unit for storing metadata of the transferred file;
A second storage unit that stores actual data satisfying a predetermined condition among the actual data of the transferred file;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, the metadata stored in the first storage unit is acquired and transmitted to the backup device, and the backup device receives the metadata. When a transmission request for the actual data of the file stored by the node is received, the actual data stored by the second storage unit and the actual storage stored by the node storing the file received by the backup device are stored. A file access unit for transmitting any of the data to the backup device;
A distributed file system comprising:
[Appendix 2]
The management device
A third storage unit that stores information for specifying a node that stores a file corresponding to the transferred file among the plurality of nodes;
When the actual data of the file that has received the transmission request from the backup device satisfies the predetermined condition, the first reference that refers to the actual data stored in the second storage unit and responds to the file access unit And
When the actual data of the file that has received the transmission request from the backup device does not satisfy the predetermined condition, the node that stores the file that has received the transmission request from the backup device with reference to the third storage unit A second reference unit that acquires actual data and outputs the actual data to the file access unit;
The distributed file system according to appendix 1, further comprising:
[Appendix 3]
When receiving the snapshot generation instruction from the backup device, the management device transmits a snapshot acquisition instruction to the control device,
The controller is
A snapshot control unit for managing the generation of the file from the time when the received snapshot acquisition instruction is received,
The snapshot control unit
After the processing of the management device relating to the old generation file among the files transferred to the management device is completed, the management device is instructed to acquire snapshots of the first and second storage units. ,
After the node completes processing related to the old generation file, it instructs the node that has completed the storage of the old generation file to acquire a snapshot,
The distributed file system according to appendix 2, wherein the management apparatus and the node notify the management apparatus that the acquisition of the snapshot has been completed after obtaining a response indicating that the acquisition of the snapshot has been completed.
[Appendix 4]
The distributed file system according to appendix 3, wherein the transfer unit transfers a new generation file to the management device after the snapshot control unit obtains a response that the acquisition of the snapshot is completed from the management device.
[Appendix 5]
The controller is
The distributed file system according to any one of appendices 1 to 4, further comprising a storage unit that temporarily stores a file to be transferred to the management apparatus.
[Appendix 6]
The management device
The distributed file system according to any one of appendices 1 to 5, further comprising a fourth storage unit that stores an update history of a file transferred from the control device.
[Appendix 7]
The distributed file system according to any one of appendices 1 to 6, wherein the predetermined condition includes at least a condition relating to a size of the transferred file.
[Appendix 8]
A control device for controlling access to a plurality of nodes storing files;
A management device that manages information required to process a request from the backup device;
A backup file acquisition method for a distributed file system comprising:
Transferring the file to be stored to the management device when storing the file in at least one of the plurality of nodes;
Storing the metadata of the transferred file in a first storage unit;
Storing actual data satisfying a predetermined condition in the second storage unit among the actual data of the transferred file;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, obtaining metadata stored in the first storage unit and transmitting the metadata to the backup device;
A node that stores the actual data stored in the second storage unit and the file that has received the transmission request from the backup device when receiving a transmission request for the actual data of the file stored in the node from the backup device Transmitting any of the actual data stored in the backup device,
Including backup file acquisition method.
[Appendix 9]
Control access to multiple nodes that store files,
A control device that transfers a file to be stored to a management device that manages information required to process a request from a backup device when storing the file in at least one of the plurality of nodes. .
[Appendix 10]
A management device that manages information required to process a request from a backup device,
A first storage unit that stores metadata of a file transferred from a control device that controls access to a plurality of nodes that store the file;
A second storage unit that stores actual data satisfying a predetermined condition among the actual data of the file transferred from the control device;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, the metadata stored in the first storage unit is acquired and transmitted to the backup device, and the backup device receives the metadata. When a transmission request for the actual data of the file stored by the node is received, the actual data stored by the second storage unit and the actual storage stored by the node storing the file received by the backup device are stored. A file access unit for transmitting any of the data to the backup device;
A management device comprising:
[Appendix 11]
A program to be executed by a computer that controls a control device that controls access to a plurality of nodes that store files,
When a file is stored in at least one of the plurality of nodes, a process of transferring the file to be stored to a management device that manages information required for processing a request from a backup device is executed. Program to make.
[Appendix 12]
A program to be executed by a computer that controls a management device that manages information required when processing a request from a backup device,
Processing for storing, in the first storage unit, metadata of a file transferred from a control device that controls access to a plurality of nodes that store the file;
A process of storing actual data satisfying a predetermined condition among the actual data of the file transferred from the control device in the second storage unit;
When receiving a transmission request for metadata related to a file stored in the node from the backup device, a process of acquiring metadata stored in the first storage unit and transmitting the metadata to the backup device;
A node that stores the actual data stored in the second storage unit and the file that has received the transmission request from the backup device when receiving a transmission request for the actual data of the file stored in the node from the backup device Processing for transmitting any of the actual data stored in the backup device,
A program that executes
Note that the programs of Supplementary Note 11 and Supplementary Note 12 can be recorded on a computer-readable storage medium. The storage medium may be non-transient such as a semiconductor memory, a hard disk, a magnetic recording medium, an optical recording medium, or the like.

  Each form of appendix 8 to appendix 12 can be developed into the form of appendix 2 to appendix 7 in the same manner as appendix 1.

  Each disclosure of the cited patent documents and the like cited above is incorporated herein by reference. Within the scope of the entire disclosure (including claims) of the present invention, the embodiments and examples can be changed and adjusted based on the basic technical concept. In addition, various combinations or selections of various disclosed elements (including each element in each claim, each element in each embodiment or example, each element in each drawing, etc.) within the scope of the entire disclosure of the present invention. Is possible. That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the entire disclosure including the claims and the technical idea. In particular, with respect to the numerical ranges described in this document, any numerical value or small range included in the range should be construed as being specifically described even if there is no specific description.

1, 60 Client 2, 2a-2c, 61, 64 File system integration device 3, 3-1 to 3-n, 3a to 3i, 62-1 to 62-6 Distributed node 4, 4a to 4c, 63, 64 Backup Apparatus 5, 5a Replicated file system management apparatus 10, 10a, 10b, 30, 300, 400 File access unit 11, 11a, 11b, 301, 401 Distributed storage unit 12 File transfer unit 13 Snapshot control unit 14 Update information storage unit 20 20a, 20b, 51, 51a, 302 Data storage unit 21, 21a, 21b, 303 Metadata storage unit 31 Snapshot unit 32, 32a Replicated file system unit 50, 50a Block storage control unit 52 File transfer unit 65 Metadata storage Device 100 control device 101 management device 102 transfer unit 103 first Storage unit 104 Second storage unit 105 File access unit 201 Indirect data storage unit 202 Indirect data reference unit 203 Transfer actual data storage unit 204 Actual data reference unit 205 Transfer metadata storage unit 206 File update history storage unit 402 Metadata management unit

Claims (10)

A control device for controlling access to a plurality of nodes storing files;
A management device that manages information required to process a request from the backup device;
Including
The controller is
A transfer unit that transfers the file to be stored to the management device when storing the file in at least one of the plurality of nodes;
The management device
A first storage unit for storing metadata of the transferred file;
A second storage unit that stores actual data satisfying a predetermined condition among the actual data of the transferred file;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, the metadata stored in the first storage unit is acquired and transmitted to the backup device, and the backup device receives the metadata. When a transmission request for the actual data of the file stored by the node is received, the actual data stored by the second storage unit and the actual storage stored by the node storing the file received by the backup device are stored. A file access unit for transmitting any of the data to the backup device;
A distributed file system comprising: The management device
A third storage unit that stores information for specifying a node that stores a file corresponding to the transferred file among the plurality of nodes;
When the actual data of the file that has received the transmission request from the backup device satisfies the predetermined condition, the first reference that refers to the actual data stored in the second storage unit and responds to the file access unit And
When the actual data of the file that has received the transmission request from the backup device does not satisfy the predetermined condition, the node that stores the file that has received the transmission request from the backup device with reference to the third storage unit A second reference unit that acquires actual data and outputs the actual data to the file access unit;
The distributed file system of claim 1, further comprising: When receiving the snapshot generation instruction from the backup device, the management device transmits a snapshot acquisition instruction to the control device,
The controller is
A snapshot control unit for managing the generation of the file from the time when the received snapshot acquisition instruction is received,
The snapshot control unit
After the processing of the management device relating to the old generation file among the files transferred to the management device is completed, the management device is instructed to acquire snapshots of the first and second storage units. ,
After the node completes processing related to the old generation file, it instructs the node that has completed the storage of the old generation file to acquire a snapshot,
3. The distributed file system according to claim 2, wherein after obtaining a response indicating that snapshot acquisition has been completed from the management device and the node, the management device and the node are notified of the completion of snapshot acquisition.   The distributed file system according to claim 3, wherein the transfer unit transfers a new generation file to the management device after the snapshot control unit obtains a response that the acquisition of the snapshot is completed from the management device. The controller is
The distributed file system according to claim 1, further comprising a storage unit that temporarily stores files to be transferred to the management apparatus. The management device
The distributed file system according to claim 1, further comprising a fourth storage unit that stores an update history of a file transferred from the control device.   The distributed file system according to claim 1, wherein the predetermined condition includes at least a condition related to a size of the transferred file. A control device for controlling access to a plurality of nodes storing files;
A management device that manages information required to process a request from the backup device;
A backup file acquisition method for a distributed file system comprising:
Transferring the file to be stored to the management device when storing the file in at least one of the plurality of nodes;
Storing the metadata of the transferred file in a first storage unit;
Storing actual data satisfying a predetermined condition in the second storage unit among the actual data of the transferred file;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, obtaining metadata stored in the first storage unit and transmitting the metadata to the backup device;
A node that stores the actual data stored in the second storage unit and the file that has received the transmission request from the backup device when receiving a transmission request for the actual data of the file stored in the node from the backup device Transmitting any of the actual data stored in the backup device,
Including backup file acquisition method. Control access to multiple nodes that store files,
A control device that transfers a file to be stored to a management device that manages information required to process a request from a backup device when storing the file in at least one of the plurality of nodes. . A management device that manages information required to process a request from a backup device,
A first storage unit that stores metadata of a file transferred from a control device that controls access to a plurality of nodes that store the file;
A second storage unit that stores actual data satisfying a predetermined condition among the actual data of the file transferred from the control device;
When receiving a transmission request for metadata relating to a file stored in the node from the backup device, the metadata stored in the first storage unit is acquired and transmitted to the backup device, and the backup device receives the metadata. When a transmission request for the actual data of the file stored by the node is received, the actual data stored by the second storage unit and the actual storage stored by the node storing the file received by the backup device are stored. A file access unit for transmitting any of the data to the backup device;
A management device comprising:

Images