JP2012053668A - File management program, file management method and file management device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology for suppressing the generation of a fragment in a network file system.SOLUTION: A series of file write-in requests with respect to the same file to be transmitted from a client 20 to an NFS server 10 include the identification information of the same file. In the NFS server 10, a request acceptance part 11 accepts the file write-in request from the client 20, and a determination information storage part 125 stores determination information in which the association of the identification information of the file with a thread 13 which has processed the file write-in request including the identification information of the file in the past is recorded, and a processing thread assignment part 12 assigns all the processing of the series of file write-in requests including the identification information of the same file to the same thread 13 by referring to the determination information.

Description

  The present invention relates to a file management program, a file management method, and a file management apparatus for managing files that other apparatuses access via a network in a network file system.

  The network file system is a file system that makes it possible to use a file on a remote computer as if it were a local file on the local computer. Here, a computer that accesses a file in another computer is called a client, and a computer that manages a file accessed by the client is called a file server.

  In a multiprocessor configuration device, by transmitting thread identifiers through inter-processor communication, it is possible to achieve consistent processing for each processor and prevent re-transmission of information required for processing. Technology is known. In addition, a file server technique for processing a request from a client with multithread is known.

JP-A-6-348558 Japanese Patent Laid-Open No. 8-212090

  In a file server that performs multithread processing in a network file system, processing for a request from a client is allocated to one of the threads. The thread that processes the request from the client accesses the disk by the local file system of the file server.

  In the client, when the size of data to be written is large, the file write request is divided into a plurality of requests and transmitted to the file server. In the file server, multiple file write requests for the same file received from the same client are allocated among multiple threads. Each thread to which processing is allocated performs write processing to the local file system. At this time, fragmentation may occur on the disk. Fragment is the fragmentation of a file on a storage device such as a disk.

  In a file server, a thread that performs a write process on the same file needs to acquire a lock. The thread that acquired the lock performs the process of writing the file to the local file system. Other threads that could not acquire the lock will wait for the file writing process of the thread that acquired the lock to end. When the file writing process by the thread that acquired the lock is completed, the thread that has been waiting acquires the lock.

  At this time, the thread that processes the requests in the order in which the divided write requests are transmitted from the client may not always acquire the lock. Therefore, multiple write requests for the same file may be processed out of order and fragmentation may occur. When a fragment occurs, the file cannot be read continuously from the area on the disk when the file is read, and a disk seek occurs and the file reading performance deteriorates.

  One aspect of the disclosed technology aims to solve the above problems and provide a technology for suppressing the occurrence of fragments in a network file system.

  The disclosed program causes a computer that manages files that other devices access via a network to function as follows.

  In other words, the program is installed and executed, and the computer includes a storage unit that stores determination information corresponding to the thread and file identification information. If there is a thread corresponding to the file identification information included in the file write request in the determination information read from the file, the thread corresponding to the file identification information included in the file write request is designated as a thread that processes the file write request. And if there is no thread corresponding to the identification information of the file included in the file write request in the determination information, a thread that is not associated in the determination information is set as a thread for processing the file write request, The correspondence between the identification information of the file included in the thread and the file write request for performing management, to execute a step of recording the determination information.

  With the disclosed technology, it is possible to suppress the occurrence of fragments in the network file system. As a result, it is possible to prevent a decrease in file reading performance in the network file system.

It is a figure explaining the example of the existing network file system. It is a figure explaining the example of fragment generation. It is a figure which shows the structural example of the network file system by this Embodiment. It is a figure which shows the structural example of the computer which implement | achieves the NFS server by this Embodiment. It is a figure which shows the example of the data structure of the file write request by this Embodiment. It is a figure which shows the example of the determination information by this Embodiment. It is a figure explaining allocation of the logical block number by the network file system of this Embodiment. It is a file management processing flowchart (1) by the NFS server of this Embodiment. It is a file management processing flowchart (1) by the NFS server of this Embodiment. It is a file management processing flowchart (2) by the NFS server of this Embodiment. It is a file management processing flowchart (2) by the NFS server of this Embodiment. It is a figure explaining the effect of the network file system by this Embodiment.

  Hereinafter, the present embodiment will be described with reference to the drawings.

  FIG. 1 is a diagram for explaining an example of an existing network file system.

  In the network file system, files on other devices can be used as if they were local to the device. A typical network file system is NFS (Network File System) of Sun Microsystems. NFS is a trademark of Sun Microsystems, Inc. in the United States.

  The network file system shown in FIG. 1 is an example of an existing general client / server type network file system. In the network file system shown in FIG. 1, an NFS server 510 is a device that manages files accessed by other devices. The NFS server 510 corresponds to the file server described above. The client 520 is a device that uses a file managed by the NFS server 510. The client 520 accesses a file on the NFS server 510 via a network 530 such as the Internet or a LAN (Local Area Network).

  An example in which the client 520 writes a file to the NFS server 510 will be described with reference to FIG.

  The client 520 transmits a file write request for file writing to the NFS server 510. At this time, the write request for the same file may be divided into a plurality of requests and transmitted depending on the size of the data to be written and the setting of the network file system. Here, as shown in FIG. 1, in the client 520, the write request for the same file is divided into three requests of request # 1, request # 2, and request # 3 in order from the top of the data to be written. Assume that it is sent to the server 510.

  The NFS server 510 receives a request from the client 520. In the NFS server 510, the processing thread allocation unit 511 allocates the processing of the request received by the NFS server 510 to any thread 512 in the idle state. The NFS server 510 has a plurality of threads 512 (a, b, c,...). Each thread performs processing for the request allocated by the processing thread allocation unit 511. The thread that has finished processing enters the idle state and waits for the next processing to be allocated.

In the example illustrated in FIG. 1, three file write requests transmitted from the client 520 are received by the NFS server 510 via the network 530. In the NFS server 510, the processing thread allocation unit 511 allocates the processing of the three file write requests received sequentially to each thread in the idle state. Here, the processing of request # 1 is allocated to thread (th ₀ ) 512a, the processing of request # 2 is allocated to thread (th ₁ ) 512b, and the processing of request # 3 is allocated to thread (th ₂ ) 512c. The parentheses for each thread 512 indicate the identification information of each thread.

  When the processing of the received request is allocated, the thread 512 interprets the request and executes the processing requested by the request.

  The thread 512 to which file write request processing is assigned performs file write to the local file system 513. At this time, the thread 512 that performs the writing process of the divided same file needs to acquire the lock. The thread 512 that has acquired the lock performs a file writing process to the local file system 513. The other thread 512 that could not acquire the lock waits for the end of the file writing process of the thread 512 that acquired the lock. When the file writing process by the thread 512 that has acquired the lock is completed, the thread 512 that has been in the lock acquisition waiting state acquires the lock.

  The local file system 513 is a file control mechanism in the OS of the NFS server 510. The local file system 513 assigns a logical block number that is an arrangement destination of file data on the disk. The logical block number is information indicating an address on the disk. At this time, the local file system 513 attempts to assign a logical block number as close as possible to the logical block number assigned last when writing to the same file. In the network file system shown in FIG. 1, the file storage device 514 is a disk on which a file is written.

In the network file system shown in FIG. 1, th ₀ , th ₁ , and th ₂ threads 512 to which processing for a divided write request for the same file is allocated respectively attempt to acquire a lock.

  FIG. 2 is a diagram illustrating an example of fragment generation.

  The example shown in FIG. 2 is an example in which a fragment is generated on the disk of the file storage device 514 in the example of the network file system shown in FIG.

As shown in FIG. 2, it is assumed that the thread (th ₀ ) 512a that first processes the request # 1 acquires a lock and performs a file write process to the local file system 513. At this time, the thread (th ₁ ) 512b that performs the processing of request # 2 and the thread (th ₂ ) 512c that performs the processing of request # 3 wait until the writing processing by the thread (th ₀ ) 512a is completed. .

The local file system 513 assigns a logical block number 12345 to the file data of request # 1 processed by the thread (th ₀ ) 512a. When the processing of the thread (th ₀ ) 512a is finished, the thread 512 waiting for lock acquisition acquires the lock and performs processing.

At this time, the thread (th ₁ ) 512b that performs the processing of the request # 2 does not necessarily acquire the next lock. Here, as shown in FIG. 2, it is assumed that the thread (th ₂ ) 512c that performs the processing of the request # 3 acquires the lock first.

The file data of the request # 3 processed by the thread (th ₂ ) 512c is divided from the same file as the file data of the previous request # 1. The local file system 513 assigns a logical block number 12346 continuous with the logical block number 12345 at the time of the previous request # 1 processing to the file data of the request # 3 processed by the thread (th ₂ ) 512c.

When the processing of the thread (th ₂ ) 512c is completed, the thread (th ₁ ) 512b in the lock acquisition waiting state finally acquires the lock and performs processing. The file data of the request # 2 processed by the thread (th ₁ ) 512b is divided from the same file as the file data of the previous request # 3. The local file system 513 assigns a logical block number 12347 continuous with the logical block number 12346 at the time of the previous request # 3 processing to the file data of the request # 2 processed by the thread (th ₁ ) 512b.

  As a result, as shown in FIGS. 1 and 2, in the file storage device 514, the file data is written in the order of request # 1 → request # 3 → request # 2 on the continuous disk area of the logical block numbers 12345 to 12347. It is. Since the actual data order in the file is in the order of request # 1 → request # 2 → request # 3, fragmentation occurs on the disk.

  When this file is read, the data is read in the order of logical block numbers 12345 → 12347 → 12346. That is, since continuous reading is not performed, a disk seek occurs and the file reading performance deteriorates.

  Such a problem becomes particularly noticeable when a broadband network such as IPoIB (QDR) (Internet Protocol over InfiniBand (Quad Data Rate)) is used as the network 530. InfiniBand is a registered trademark of InfiniBand Trade Association.

  When a broadband IPoIB (QDR) or the like is used as the network 530, the number of requests received per hour to the NFS server 510 is larger than when a GbE (Gigabit Ethernet) or the like is used as the network 530. . Therefore, when broadband IPoIB (QDR) or the like is used as the network 530, there is a high possibility that a plurality of threads 512 that perform file writing processing exist simultaneously in the NFS server 510. At this time, as described above, the thread 512 that writes data later in the same file acquires the lock first, and the possibility of fragmentation increases.

  In the following, a network file system according to this embodiment for suppressing the occurrence of such a fragment will be described.

  FIG. 3 is a diagram showing a configuration example of the network file system according to the present embodiment.

  The network file system shown in FIG. 3 includes an NFS server 10, a plurality of clients 20 (a, b, c,...), And a network 30. In the network file system shown in FIG. 3, the NFS server 10 is a device that manages files accessed by other devices. The client 20 is a device that uses a file managed by the NFS server 10.

  As shown in FIG. 3, a plurality of clients 20 can access files on the NFS server 10 via the network 30. In the network file system shown in FIG. 3, the network 30 is a broadband network such as IPoIB (QDR).

  When writing a file to the NFS server 10, the client 20 transmits a file write request to the NFS server 10. At this time, the client 20 divides the file write request into a plurality of requests according to the size of the data to be written and the setting of the network file system. The client 20 transmits a divided file write request to the NFS server 10 in order from the top of the data to be written.

  In the present embodiment, the file write request sent from the client 20 to the NFS server 10 includes file identification information. The file identification information is information for identifying, for each file, a series of file write requests for the same file. That is, the same file identification information is included in a series of write requests for the same file.

  The file identification information may be information that can determine whether a series of file write requests are requests for the same file. For example, information originally included in the file write request such as the IP address of the client 20 or the file descriptor of the file to be written may be used as the file identification information. A plurality of file write requests transmitted continuously from the same client 20 in a short time is highly likely to be a series of file write requests for the same file. Also, new information may be defined for the request for file identification information. For example, a member variable that holds file identification information may be added to the request structure, and information that is considered unique for each file may be stored as file identification information.

  In the network file system shown in FIG. 3, the NFS server 10 includes a request reception unit 11, a processing thread allocation unit 12, a plurality of threads 13 (a, b, c,...), A local file system 14, and a file storage device 15. Have

  The request reception unit 11 receives a file write request transmitted from the client 20.

  The processing thread allocation unit 12 allocates the file write request received by the request reception unit 11 to the thread 13 in the idle state. At this time, the processing thread allocation unit 12 according to the present embodiment allocates a series of file write requests for the same file to the same thread 13. The processing thread allocation unit 12 includes a determination unit 121, a processing thread determination unit 122, a deletion unit 123, and a determination information storage unit 125.

  The determination information storage unit 125 is a computer-accessible storage unit that stores determination information in which correspondence between a file identification information and a thread 13 that processes a file write request including the file identification information is recorded. .

  The determination unit 121 extracts file identification information from the received file write request. The determination unit 121 refers to the determination information stored in the determination information storage unit 125 and determines whether the correspondence between the extracted file identification information and any one of the threads 13 is recorded in the determination information. If there is a record in the determination information, the determination unit 121 determines that the received file write request is a request for the same file as the file write request to which the thread 13 was previously allocated. If there is no record in the determination information, the determination unit 121 determines that the received file write request is the first request in a series of file write requests for the same file.

  The processing thread determination unit 122 determines the thread 13 to which the received file write request processing is allocated according to the determination result of the determination unit 121. When the determination information includes a record of the file identification information extracted from the file write request, the processing thread determination unit 122 writes the thread 13 associated with the file identification information extracted by the determination information to the file write The thread 13 that processes the request is determined. If the determination information does not contain the record of the file identification information extracted from the file write request, the processing thread determination unit 122 sends the file write request to the thread 13 that is not associated with the file identification information in the determination information. The thread 13 to be processed is determined.

  The processing thread determination unit 122 records the correspondence between the file identification information extracted from the file write request and the determined thread 13 for processing the file write request in the determination information stored in the determination information storage unit 125. To do. As a result, the correspondence between the file identification information and the thread 13 to which the file write request processing including the file identification information is allocated is recorded in the determination information.

  When it is determined that a series of file write requests for one and the same file have been completed, the deletion unit 123 records the identification information of the file of the file write request and the determination information stored in the determination information storage unit 125 Delete from.

  For example, when it is determined that the received file write request is the final request, the deletion unit 123 deletes the record of the identification information of the file included in the file write request from the determination information. Processing of the final request of a series of file write requests includes file close processing.

  Further, the deletion unit 123 deletes the record of the identification information of the file associated with the thread 13 for which a predetermined time or more has elapsed since the last file write request was allocated in the determination information. A series of file write requests for one and the same file is considered to be continuously transmitted from the client 20 in a short time.

  Each thread 13, local file system 14, and file storage device 15 to which file write request processing is assigned are the same as the thread 512, local file system 513, and file storage device 514 shown in FIG. Is omitted.

  FIG. 4 is a diagram illustrating a configuration example of a computer that realizes the NFS server according to the present embodiment.

  A computer 100 that implements the NFS server 10 includes a CPU (Central Processing Unit) 101, a memory 102 such as a RAM (Random Access Memory), a storage device 103 such as an HDD (Hard Disk Drive), an input device 104, an output device 105, a communication The apparatus 106 etc. are provided.

  The NFS server 10 and each functional unit included in the NFS server 10 illustrated in FIG. 3 can be realized by hardware such as the CPU 101 and the memory 102 included in the computer 100 and a software program. A program that can be executed by the computer 100 is stored in the storage device 103, read into the memory 102 at the time of execution, and executed by the CPU 101.

  The computer 100 can also read a program directly from a portable recording medium and execute processing according to the program. In addition, each time the program is transferred from the server computer, the computer 100 can sequentially execute processing according to the received program. Further, this program can be recorded on a recording medium readable by the computer 100.

  In the present embodiment, the computer that implements the client 20 also has the same hardware as the computer 100 shown in FIG.

  FIG. 5 is a diagram illustrating an example of a data structure of a file write request according to the present embodiment.

  The file write request transmitted from the client 20 has a header part and a data part. The header part has information such as the IP address of the client 20 that is the transmission source and the file descriptor of the file to be written. The data part is write data. When the request is divided into a plurality of parts, the data part becomes a part of the data to be written.

  For example, in the file write request shown in FIG. 5, information such as the IP address and file descriptor of the client 20 included in the header part can be file identification information used for determining a series of file write requests for the same file.

  FIG. 6 is a diagram illustrating an example of determination information according to the present embodiment.

  The determination information 126 illustrated in FIG. 6 is an example of information stored in the determination information storage unit 125. The determination information 126 includes thread, file identification information, and time information.

  In the determination information 126 illustrated in FIG. 6, the thread indicates identification information of each thread 13 that operates on the NFS server 10.

  In the determination information 126 illustrated in FIG. 6, a record in which the file identification information is recorded is a record in which the thread 13 and the file identification information are associated with each other. The thread 13 associated with the file identification information is a thread 13 that has processed a file write request including the file identification information in the past.

  In the determination information 126 shown in FIG. 6, the time is the time when the file write request process including the recorded file identification information is finally allocated to the thread 13.

  Here, an example of file writing by the network file system of the present embodiment shown in FIG. 3 will be described.

  For example, in the client 20a, a write request for the same file is divided into three requests of request # 1, request # 2, and request # 3 in order from the top of the data to be written and transmitted to the NFS server 10. . The client 20a transmits three file write requests to the NFS server 10 via the network 30 in order from the top of the file.

  In the NFS server 10, the request reception unit 11 first receives the request # 1 received first.

In the processing thread allocation unit 12, the determination unit 121 extracts file identification information from the received request # 1. Here, it is assumed that the IP address of the transmission source client 20a is extracted as file identification information. For example, the file identification information id ₀₁ is extracted from the request # 1.

The determination unit 121 determines whether or not there is a record of the extracted file identification information id _{01 in} the determination information 126 stored in the determination information storage unit 125. Here, since the request # 1 is the first request of a series of write requests for the same file, the file identification information id ₀₁ extracted in the determination information 126 is not recorded.

The processing thread determination unit 122 refers to the determination information 126 stored in the determination information storage unit 125 and selects one thread 13 that has no associated file identification information. Here, it is assumed that the th ₀ thread 13a is selected. The processing thread determination unit 122 records the identification information id ₀₁ of the file extracted from the request # 1 in the record of the th ₀ thread 13a of the determination information 126 stored in the determination information storage unit 125.

The processing thread allocation unit 12 assigns the processing of request # 1 to the th ₀ thread 13a. The thread 13a starts processing for the request # 1.

After request # 1, request acceptance unit 11 accepts request # 2. The determination unit 121 extracts the file identification information id ₀₁ from the request # 2. Since the request # 1 and the request # 2 are requests transmitted from the same client 20a, the file identification information is the same id ₀₁ .

The determination unit 121 determines whether or not there is a record of the extracted file identification information id _{01 in} the determination information 126 stored in the determination information storage unit 125. Since the process of request # 1, which is the first file write request, has already been performed among a series of file write requests including request # 2, the determination information 126 includes a record of file identification information id ₀₁ . Processing thread decision unit 122 refers to the determination information 126 stored in the determination information storage unit 125, as a thread 13 corresponding to the identification information id ₀₁ files, extracts the threads 13a of th _0.

If the extracted th ₀ thread 13a is not being processed, the processing thread allocation unit 12 allocates the processing of request # 2 to the th ₀ thread 13a. The thread 13a starts processing for the request # 2.

It should be noted that the processing thread allocating unit 12 makes a request to the thread 13 even if the th ₁ thread 13b and the th ₂ thread 13c are in the idle state when the extracted th ₀ thread 13a is being processed. # 2 is not allocated. If the extracted th ₀ thread 13a is being processed, the NFS server 10 makes a busy response to the client 20a that is the transmission source of the request # 2. At this time, the NFS server 10 waits until, for example, the th ₀ thread 13a enters the idle state, and notifies the client 20a of the fact when the thread enters the idle state. The client 20a retransmits request # 2. In the NFS server 10, the above-described processing at the time of request reception is performed again, and the processing of request # 2 is allocated to the th ₀ thread 13 a that has entered the idle state.

  The request # 3 transmitted after the request # 2 is also processed in the same manner as the request # 2.

  FIG. 7 is a diagram for explaining logical block number assignment by the network file system according to the present embodiment.

As shown in FIG. 7, in the NFS server 10 of the present embodiment, a series of file write requests for the same file are all processed by th ₀ thread to which the process of request # 1, which is the first request, is allocated. 13a. The processes for request # 2 and request # 3 other than the first request # 1 are allocated to the same th ₀ thread 13a and executed after the processing of the previous request is completed. Since simultaneous writing by a plurality of threads 13 is not performed, the order of requests to be processed does not change depending on the lock acquisition order. That is, the order of file writing to the local file system 14 is the order of requests transmitted from the client 20a.

First, the processing thread allocation unit 12 allocates the processing of request # 1 to the th ₀ thread 13a. The thread 13a that performs processing of the allocated request # 1 performs file writing processing on the local file system 14. The local file system 14 assigns a logical block number 12345 to the file data of the request # 1 processed by the thread 13a.

Next, the processing of the request # 2 is allocated by the processing thread allocation unit 12 to the th ₀ thread 13a that has completed the processing of the request # 1. The thread 13a that performs processing of the allocated request # 2 performs file writing processing to the local file system 14. The local file system 14 assigns a logical block number 12346 to the file data of the request # 2 processed by the thread 13a.

Further, the processing of the request # 3 is allocated by the processing thread allocation unit 12 to the th ₀ thread 13a that has completed the processing of the request # 2. The thread 13a that performs processing of the allocated request # 3 performs file writing processing to the local file system 14. The local file system 14 assigns a logical block number 12347 to the file data of the request # 3 processed by the thread 13a.

  Thus, in the NFS server 10 according to the present embodiment, processing for a series of file write requests for one file is allocated to the same thread 13. Therefore, as shown in FIG. 7, in the file storage device 15, the file data of each request is written in the transmission order from the client 20a on the continuous disk area of the logical block numbers 12345 to 12347. Since the actual order of data in the file matches the order in which the file data is arranged on the disk, no fragmentation occurs on the disk.

  When this file is read, data is read in the order of consecutive logical block numbers 12345 → 12346 → 12347, so that unnecessary disk seek does not occur and the file reading performance is improved.

  8 and 9 are flowcharts (1) of file management processing by the NFS server of this embodiment.

  The flowcharts shown in FIGS. 8 and 9 show an example of the processing flow from the reception of the file write request to the allocation of the received file write request to the thread 13 in the NFS server 10. In the flowcharts shown in FIGS. 8 and 9, the record of the identification information of the corresponding file is deleted from the determination information 126 when the final request of a series of file write requests is detected.

  In the NFS server 10, the request reception unit 11 receives a file write request from the client 20 (step S10).

  In the processing thread allocation unit 12, the determination unit 121 extracts file identification information from the received file write request (step S11). The determination unit 121 determines whether there is a record of the extracted file identification information in the determination information 126 stored in the determination information storage unit 125 (step S12).

  If the identification information 126 of the extracted file is recorded in the determination information 126 (YES in step S12), the processing thread determination unit 122 extracts the thread 13 corresponding to the extracted file identification information from the determination information 126. (Step S13). The processing thread allocation unit 12 determines whether the extracted thread 13 is being processed (step S14).

  If the extracted thread 13 is being processed (YES in step S14), the NFS server 10 returns a busy response to the client 20 that has transmitted the file write request (step S15).

  If the extracted thread 13 is not being processed (NO in step S14), the processing thread determination unit 122 determines the extracted thread 13 to be the thread 13 that processes the received file write request (step S16). .

  The deletion unit 123 determines whether the received file write request is a final request in a series of file write requests (step S17). Processing of the final request of a series of file write requests includes file close processing. When the received file write request is the final request (YES in step S17), the deletion unit 123 uses the determination information 126 stored in the determination information storage unit 125 to determine the file extracted from the received file write request. The record of identification information is deleted (step S18).

  The processing thread allocation unit 12 allocates the received file write request processing to the determined thread 13 (step S19).

  If the identification information 126 of the extracted file is not recorded in the determination information 126 (NO in step S12), the processing thread determination unit 122 determines whether there is a thread 13 in the determination information 126 that is not associated with the file identification information. (Step S20).

  If there is no thread 13 that is not associated with file identification information (NO in step S20), the NFS server 10 returns a busy response to the client 20 that sent the file write request (step S21).

  If there is a thread 13 that is not associated with file identification information (YES in step S20), the processing thread determination unit 122 selects one thread 13 that is not associated with file identification information from the determination information 126 ( Step S22). The processing thread determination unit 122 determines the selected thread 13 as the thread 13 that processes the received file write request (step S23).

  The processing thread determination unit 122 determines whether the received file write request is a final request (step S24). If the received file write request is not the final request (NO in step S24), the processing thread determination unit 122 records the correspondence between the determined thread 13 and the extracted file identification information in the determination information 126. (Step S25).

  The processing thread allocation unit 12 allocates the received file write request processing to the determined thread 13 (step S26).

  10 and 11 are flowcharts (2) of the file management process by the NFS server according to the present embodiment.

  The flowcharts shown in FIGS. 10 and 11 show an example of the flow of processing from the reception of a file write request to the allocation of the received file write request to the thread 13 in the NFS server 10. In the flowcharts shown in FIGS. 10 and 11, the recording of the file identification information of the determination information 126 is invalidated for the thread 13 for which a predetermined time has elapsed since the last allocation.

  In the NFS server 10, the request reception unit 11 receives a file write request from the client 20 (step S30).

  In the processing thread allocation unit 12, the determination unit 121 extracts file identification information from the received file write request (step S31). The determination unit 121 determines whether there is a record of the extracted file identification information in the determination information 126 stored in the determination information storage unit 125 (step S32).

  If the identification information 126 of the extracted file is recorded in the determination information 126 (YES in step S32), the processing thread determination unit 122 extracts the thread 13 corresponding to the extracted file identification information from the determination information 126. (Step S33). The processing thread allocation unit 12 determines whether the extracted thread 13 is being processed (step S34).

  If the extracted thread 13 is being processed (YES in step S34), the NFS server 10 returns a busy response to the client 20 that has transmitted the file write request (step S35).

  If the extracted thread 13 is not being processed (NO in step S34), the processing thread determination unit 122 determines the extracted thread 13 as the thread 13 that processes the received file write request (step S36). . The processing thread determination unit 122 updates the time of the determined thread 13 in the determination information 126 (step S37).

  The processing thread allocation unit 12 allocates the received file write request processing to the determined thread 13 (step S38).

  If the identification information 126 of the extracted file is not recorded in the determination information 126 (NO in step S32), the processing thread determination unit 122 determines whether the determination information 126 includes a thread 13 that does not correspond to the file identification information. (Step S39).

  If there is a thread 13 that is not associated with file identification information (YES in step S39), the processing thread determination unit 122 selects one thread 13 that is not associated with file identification information from the determination information 126 ( Step S40).

  The processing thread determination unit 122 determines the selected thread 13 as the thread 13 that processes the received file write request (step S41). The processing thread determination unit 122 records the correspondence between the determined thread 13 and the extracted file identification information in the determination information 126 (step S42). The processing thread determination unit 122 records the determined time of the thread 13 in the determination information 126 (step S43).

  The processing thread allocation unit 12 allocates the received file write request processing to the determined thread 13 (step S44).

  If there is no thread 13 that is not associated with file identification information (NO in step S39), the processing thread determination unit 122 determines whether there is a thread 13 for which a predetermined time has elapsed since the last allocation to the determination information 126. Is determined (step S45).

  If there is no thread 13 for which the predetermined time has passed in the determination information 126 (NO in step S45), the NFS server 10 returns a busy response to the client 20 that has transmitted the file write request (step S46).

  If there is a thread 13 for which the predetermined time has elapsed in the determination information 126 (YES in step S45), the processing thread determination unit 122 selects one thread 13 for which the predetermined time has elapsed from the determination information 126 (step S47). Thereafter, the above-described processing of step S41 to step S44 is performed.

  As described above, the flowcharts shown in FIGS. 10 and 11 show an example in which determination information 126 determines that recording of file identification information is invalid after a predetermined time has elapsed since the last allocation. ing. In addition to the processing for the request, the deletion unit 123 deletes the record of the identification information of the file associated with the thread 13 that has passed a predetermined time or more after the last allocation in the determination information 126. May be. In this case, the processing in step S46 and step S47 is not necessary.

  FIG. 12 is a diagram for explaining the effect of the network file system according to this embodiment.

  FIG. 12 shows the results of experiments performed under the same hardware configuration and the same conditions in the existing network file system shown in FIG. 1 and the network file system of the present embodiment shown in FIG. It is shown. The difference between the existing network file system and the network file system of the present embodiment is the processing thread allocation mechanism. IPoIB (QDR) is adopted for the network.

  FIG. 12A shows a state of fragmentation when a 1 GB file is created from the client to the NFS server. In FIG. 12A, extents is a numerical value that represents how many files are divided on the disk. The larger the value of extents, the more fragments are generated. From the experimental results shown in FIG. 12A, it can be seen that in the network file system of this embodiment, the generation of fragments is greatly suppressed as compared with the existing network file system.

  FIG. 12B shows the reading speed when the client reads a 1 GB file created on the NFS server. From the experimental results shown in FIG. 12B, it can be seen that in the network file system of this embodiment, the reading speed for the file created in the NFS server is about three times faster than the existing network file system.

  As described above, the network file system according to the present embodiment makes it possible to suppress fragmentation of a file when writing the file on the NFS server 10 and prevent a decrease in the reading speed for the file that has been written. It becomes possible to do.

  Although the present embodiment has been described above, the present invention can naturally be modified in various ways within the scope of the gist thereof.

DESCRIPTION OF SYMBOLS 10 NFS server 11 Request reception part 12 Processing thread allocation part 121 Judgment part 122 Processing thread determination part 123 Deletion part 125 Judgment information storage part 13 Thread 14 Local file system 15 File storage apparatus 20 Client 30 Network

Claims (3)

A program for causing a computer managing a file accessed by another device via a network,
The computer including a storage unit that stores determination information corresponding to identification information of a thread and a file,
A procedure for accepting file write requests from other devices;
If the determination information read from the storage unit includes a thread corresponding to the file identification information included in the file write request, the thread corresponding to the file identification information included in the file write request The procedure to make a thread to process the request,
When the determination information does not include a thread corresponding to the identification information of the file included in the file write request, a thread that is not associated with the determination information is set as a thread for processing the file write request;
A file management program for executing a procedure for recording, in the determination information, a correspondence between a thread for performing the processing and identification information of a file included in the file write request. A file management method by a computer for managing files accessed by other devices via a network,
The computer including a storage unit that stores determination information corresponding to identification information of a thread and a file,
The process of accepting file write requests from other devices,
If the determination information read from the storage unit includes a thread corresponding to the file identification information included in the file write request, the thread corresponding to the file identification information included in the file write request The process of making a request processing thread,
When the determination information does not include a thread corresponding to the identification information of the file included in the file write request, a thread that is not associated with the determination information is set as a thread that processes the file write request;
A file management method comprising: executing a process of recording a correspondence between a thread performing the processing and file identification information included in the file write request in the determination information. A file management device that manages files that other devices access via the network,
A determination information storage unit for storing determination information corresponding to the identification information of the thread and the file;
A request reception unit for receiving file write requests from other devices;
If the determination information includes a thread corresponding to the file identification information included in the file write request, a thread corresponding to the file identification information included in the file write request is processed as the thread that processes the file write request. And if there is no thread corresponding to the identification information of the file included in the file write request in the determination information, a thread not associated with the determination information is set as a thread for processing the file write request. A file management apparatus comprising: a processing thread allocating unit that records a correspondence between a thread that performs the process and identification information of a file included in the file write request in the determination information.

Images