JP2010237837A - File system, and data rearrangement method and program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a file system, a data rearrangement method and a data rearrangement program, for rearranging data stored in a disk device, to increase data access speed even when a plurality of processes are executed in parallel. <P>SOLUTION: A data rearrangement means 124 rearranges the data stored in a storage area of the disk device 2, so that the data are consecutive according to an order read into an application 11, based on an access pattern table indicating a list of the data read into a file management means 121 in response to a request from the application. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a file system that manages access to data stored in a disk device, a data relocation method thereof, and a data relocation program.

  When a file system in a computer system repeatedly rewrites or deletes a file on a disk device in response to a request from an application, the file stored in the disk device is fragmented, and one file is distributed over a plurality of data in the storage area. Stored in separate locations. When this file fragmentation proceeds, the file read time becomes longer than when one file is recorded in a continuous area.

  In order to eliminate the fragmentation of the file, a normal file system has a defragmenting function for rearranging data stored in the disk device. In the normal defragmentation function, a file distributed to a plurality of data and stored in a disk device is rearranged so as to be stored in a continuous state, so that access to one file becomes sequential access. In this way, file access is speeded up.

  A technique related to this is disclosed in Patent Document 1. The technology disclosed in Patent Document 1 is based on access statistical information indicating a tendency of access that occurs on a logical unit, and determines the degree of deterioration in access performance of each file related to each access request from an application by “the number of physical divisions”. The number of accesses per unit time was calculated, and files with a large degree of degradation in access performance were preferentially rearranged.

JP 2005-284632 A

  However, in the related technology described above, the data stored in the disk device is rearranged so that it is continuously stored for each file, so that access to one file becomes sequential access and the speed of file access is increased. As shown, when accessing a plurality of files simultaneously by parallel processing, the plurality of files are alternately accessed, and high-speed file access cannot be performed.

  In recent years, computer systems equipped with a plurality of CPUs are generally known, and applications premised on executing parallel processing by a plurality of CPUs are also used. In particular, in the HPC (High Performance Computing) field, a single application is executed by a plurality of CPUs, and a large number of files are used as input data. In addition, there are applications in which all input data necessary for calculation is read into the memory immediately after execution of the application, and calculation cannot be started unless the reading is completed.

  When accessing multiple different files at the same time by parallel processing, sequential data access with normal defragmentation cannot be performed, and large amounts of data are input and output with low-speed random access. There was a problem that the execution speed of the application could not be increased.

  This problem will be specifically described with reference to FIG. 7. An application 71 executed on the host 7 shown in FIG. 7 includes a process 711, a process 712, a process 713, and a process 714 that are executed in parallel. It is configured. The process 711 reads the file 81 stored in the disk device 8 and reads data 811 by the first reading and data 812 by the first reading. The process 712 reads sequentially from the file 82 stored in the disk device 8 and reads data 822 in the first reading and data 822 in the first reading. The same applies to the processes 713 and 714. The file 81, the file 82, the file 83, and the file 84 are arranged in continuous areas on the disk device 8 by the defragmentation function in the host 7, respectively.

  In this case, as access to the disk device 8, first, access to the data 811, data 821, data 831, and data 841 is performed in parallel, but these are arranged in discontinuous areas in the disk device 8. This is not a sequential access, but a low-speed random access. In the subsequent access, access to the data 812, data 822, data 832, and data 842 is performed in parallel, but since these are also arranged in a discontinuous area in the storage area of the disk device 8, low-speed random access is performed. Access.

  In this way, the normal defragmentation technology places each file in a continuous storage area. Therefore, when multiple processes access another file in parallel, the disk device has low-speed random access, and the file is accessed. The input / output of was slow.

  In addition, as another related technique, a technique for accelerating data access by placing frequently accessed data on a high-speed disk device or main memory is known, but as in the HPC field, It was not applicable when inputting / outputting a large amount of data.

  Therefore, the present invention improves the above-mentioned problems of the related art and rearranges the file data stored in the disk device so that the file access speed is increased even when a plurality of processes are executed simultaneously by parallel processing. An object of the present invention is to provide a file system, a data relocation method, and a data relocation program that can shorten the effective time of an application.

  In order to achieve the above object, a file system of the present invention is a file system comprising file management means for executing reading / writing and deletion of a file stored in a disk device in response to a request from an application, An access pattern storage unit for storing an access pattern table indicating a list of data to be read by the file management unit in response to the request, and data stored in the disk device based on the stored access pattern table And a data rearranging means for rearranging the data on the storage area of the disk device so that they are continuously arranged in the order read by the application.

  The data rearrangement method of the present invention includes a file management unit that executes reading / writing and deletion of a file stored in the disk device in response to a request from an application, and the file management unit in response to a request from the application. And an access pattern storage unit for storing an access pattern table indicating a list of data to be read by the data relocation means, which reads the access pattern table from the access pattern storage unit, and based on the access pattern table The data rearrangement means identifies the data read by the application and the order in which it is read, and the data rearrangement means rearranges the data stored in the disk device so as to be continuous in the specified order. Specially placed To.

  The data rearrangement program according to the present invention includes a file management function for executing read / write and deletion of a file stored in the disk device in response to a request from the application, and the file management in response to a request from the application. Based on an access pattern table showing a list of data to be read by the function, a reading order specifying function for specifying data read by the application and its reading order, and storing in the disk device so as to be continuous in the specified order It is characterized by causing a computer to execute a data rearrangement function for rearranging stored data.

  Since the present invention is configured as described above, the file data stored in the storage area of the disk device is rearranged so as to be continuously arranged in the order of reading by the application. Even when accessing multiple files at the same time, the series of file accesses becomes sequential access in the disk device, so that the data read operation of the disk device can be physically accelerated. The processing speed can be effectively increased.

It is a functional block diagram which shows the structure of the file system of one Embodiment concerning this invention. It is a figure which shows an example of the access pattern table in embodiment disclosed in FIG. It is a figure which shows the arrangement | positioning state of the data stored in the disk apparatus 2 in embodiment disclosed in FIG. It is a figure which shows the other example of the access pattern table in embodiment disclosed in FIG. It is a flowchart which shows operation | movement of the file system of embodiment disclosed in FIG. It is a flowchart which shows operation | movement of the file system of embodiment disclosed in FIG. It is explanatory drawing showing file access at the time of performing a some process simultaneously with an application by parallel processing.

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

  FIG. 1 is a functional block diagram showing the configuration of the file system 12 of this embodiment. As shown in FIG. 1, in this embodiment, a host 1 and a disk device 2 are connected, and the host 1 manages a file stored in the disk device 2 in response to execution of an application 11. A system 12 is provided. The disk device 2 stores a file used for executing the application 11. The application 11 includes a plurality of processes, and the process is a unit of processing of the application 11.

  The file system 12 according to the present embodiment includes a file management unit 121 that reads / writes and deletes a file from / to the disk device 2 in response to a request from the application 11, and data that is connected to the disk device 2 to input / output data. And an input / output unit 125.

  Further, the file system 12 includes an access pattern storage unit 123 that stores an access pattern table indicating a list of data read by the file management unit 121 in response to a request from the application 11, and a disk based on the stored access pattern table. Data relocation means 124 is provided for relocating the data stored in the device 2 on the storage area of the disk device 2 so that the data stored in the device 2 is continuously arranged in the order of reading.

  Further, the file system 12 includes an access pattern recording unit 122 that sequentially records information related to data read by the file management unit 121 during execution of the application 11 in the access pattern table.

  The file management unit 121 is a unit that records, moves, and deletes data in the storage area of the disk device 2 via the data input / output unit 125 according to a preset method. The file management unit 121 of this embodiment has a function of inputting a request from the application 11 and recording, reading, updating, and deleting file data in the disk device 2 in accordance with the request.

  Here, the file management unit 121 of this embodiment divides the storage area of the disk device 2 into a plurality of blocks, manages the data contents recorded in each block, and the block size is 4096 bytes. Is set. For this reason, data is recorded in units of 4096 bytes on the disk device 2 of the present embodiment.

  The access pattern recording unit 122 is a unit that records a file reading pattern by the application 11 in the access pattern table when the application 11 is executed. The access pattern recording unit 122 acquires a data access request from the application 11 from the file management unit 121 and records the request in the access pattern table. Thereby, the actual file reading pattern of the application 11 can be recorded.

  The access pattern storage unit 123 stores an access pattern table indicating the content of data read by each process of the application 11 in the order of reading. The access pattern table shows the request contents from the application 11 to the file management means. FIG. 2 is a diagram illustrating an example of an access pattern table in the present embodiment.

  In the access pattern table shown in FIG. 2, the read file name, the read start position (offset), and the read size are recorded in order of reading for each process in the application 11. In the access pattern table shown in FIG. 2, the process A first reads data from offset 0 to 4096 bytes of file a (entry 1), and the second time reads data from offset 4096 bytes to 10240 bytes of file a ( Entry 2) It is recorded that data for 6144 bytes is read from offset 14336 bytes of file a for the third time (entry 3), and process B first reads data for offset 0 to 4096 bytes of file b. (Entry 1) Data for offset 4096 to 4096 bytes of file b is read the second time (entry 2) Data for offset 8192 to 8192 bytes of file b can be read for the third time (entry 3) Recorded To have.

  FIG. 4 is a diagram showing another example of the access pattern table in the present embodiment. The access pattern table shown in FIG. 4 records the start time of reading together with the file name, offset, and size.

  Here, in FIG. 1, the access pattern storage unit 123 is configured by a memory mounted on the host 1, but is not limited thereto, and may be configured by a part of the storage area of the disk device 2. . With this configuration, when a host different from the host 1 is connected to the disk device 2 to execute the application 11, the other host uses the access pattern table created by the host 1 in advance as a disk. The file can be read from the apparatus 2 and rearranged, and another host can read the file at high speed from the first execution of the application 11.

  The data rearrangement unit 124 rearranges the recording positions of the data stored in the disk device 2 based on the access pattern table stored in the access pattern storage unit 123 so as to be consecutive in the order in which the application 11 reads them. It is means to do.

  The data rearrangement unit 124 identifies the read data of the application 11 and the reading order based on the access pattern table stored in the access pattern storage unit 123, and arranges the disk device so as to be continuously arranged in the specified order. 2 has a function of changing the arrangement of the data stored in it. Here, the procedure for changing the arrangement of the data stored in the disk device 2 is the same as that in the normal defragmentation technique, and therefore will be omitted. Based on the access pattern table, the data rearrangement unit 124 according to the present embodiment can change the arrangement of data stored in the disk device 2 so as to be continuously arranged in the reading order when the application 11 is executed. It is a feature.

  For example, when the host 1 has a function of executing parallel processing and a plurality of processes in the application 11 are executed simultaneously, the access pattern storage unit 123 stores an access pattern table as shown in FIG. Then, the data rearrangement unit 124 calculates the total read size for each data read of the processes A and B based on the access pattern table, and the processes A and B based on the calculated totals. Any configuration having a function of specifying the order of reading each data may be used.

  The function of the data rearrangement unit 124 in this case will be specifically described. The data relocation unit 124 refers to the access pattern table as shown in FIG. 2 when the application 11 is stopped. First, for the process A, from the information that the file a of the entry 1, the offset = 0, the size = 4096 bytes. 4096-byte data starting from the offset 0 of the file a is arranged at the head of the continuous area in the disk device 2. Then, the fact that a total of 4096 bytes have been arranged for process A is stored in a memory provided in advance. Thereafter, relocation up to a total of 4096 bytes is performed for all processes other than process A.

  In this case, for the process B, 4096 bytes of data starting from the offset 0 of the file b is arranged in the second continuous area in the disk device 2 from the information that the file b of the entry 1 is offset = 0 and the size is 4096. The fact that a total of 4096 bytes have been arranged for process B is stored in the memory. As a result, relocation up to a total of 4096 bytes is completed for all processes other than process A.

  In the access pattern table, the block size in the storage area of the disk device 2 is set to 4096 bytes for the information of the file a of entry 2 for process A, offset = 4096, size = 10240, and the end is 4096 bytes. The size is rounded up from 10240 to 12288 bytes to be a unit. As a result, the 12288-byte data starting from the offset 4096 of the file a is allocated to the third of the continuous areas in the disk device 2, and the fact that a total of 16384 bytes has been allocated for the process A is stored in the memory. Thereafter, rearrangement of up to a total of 16384 bytes is performed for all processes other than process A.

  In this case, for the process B, the data of offset 4096 to 4096 bytes of the file b is arranged in the fourth continuous area of the disk device 2 from the information that the file b of entry 2 is offset 4096 and size is 4096. The fact that a total of 8192 bytes have been arranged for B is stored in the memory. Since 8192 bytes relocated for the process B is smaller than 16384 bytes relocated for the process A, the information of the disk device 2 is obtained from the information that the file b, the offset = 8192, and the size = 8192 of the entry 3 for the process B. Data of offset 8192 to 8192 bytes of file b is allocated to the fifth of the continuous areas, and the fact that a total of 16384 bytes have been rearranged for process B is stored in the memory. As a result, the rearrangement up to a total of 16384 bytes is completed for all processes other than the process A.

  For the process A in the access pattern table, the information of the file a, entry = offset = 14336, size = 6144 in entry 3 is referred to, and the block size is 4096 bytes as described above, so that the start position is in units of 4096 bytes. And the offset is rounded down to 12288. As a result, the entry 3 of process A is changed to file a, offset = 1288 and size = 8192. Since the offsets 12288 to 4096 bytes have already been arranged, the offset is changed to 16384. As a result, the entry 3 of process A is changed to file a, offset = 16384, size = 4096. Based on this information, data of offset 16384 to 4096 bytes of file a is arranged in the sixth of the continuous areas of the disk device 2. Then, the fact that a total of 20480 bytes have been arranged for the process A is stored in the memory.

  Thereafter, relocation up to a total of 20480 bytes is attempted for all processes other than process A, but nothing is done because relocation for process B is complete. For all processes other than process A, the relocation up to a total of 20480 bytes is completed. Then, search for the process that has not yet been relocated based on the access pattern table and the information on how many bytes of relocation have been completed for each process. When it is determined that the data has been completed, the data rearrangement process is terminated. The arrangement of data in the disk device 2 when rearranged in this manner is shown in FIG.

  As another example, when the access pattern storage unit 123 stores an access pattern table as shown in FIG. 4, the data rearrangement unit 124 stores data in the disk device 2 so as to be consecutive in the order of the read start time. Any configuration that rearranges stored data may be used. Specifically, the data rearrangement unit 124 refers to the access pattern table as shown in FIG. 4, and entry of the process a having a start time of 5.00 seconds and entry of a process b having a start time of 5.10 seconds. 1, an entry for process a with a start time of 6.00 seconds, an entry for process b with a start time of 6.10 seconds, an entry for process b with a start time of 7.50 seconds, and an entry for a start time of 8.50 The data stored in the disk device 2 is rearranged so that the respective data are consecutive in the order of the entry 3 of the process a in seconds.

  As a result, even when a plurality of processes perform file access at the same time when the application 11 is executed, the series of data access becomes sequential access in the disk device 2, so that the data read operation of the disk device 2 is physically fast. As a result, the processing speed of the application 11 can be increased.

  Next, the operation of the file system 12 of this embodiment will be described. Here, the following operation description is an embodiment of the data rearrangement method of the present invention.

  FIG. 5 is a flowchart showing an access pattern recording operation in the file system 12 of this embodiment. As shown in FIG. 5, in the access pattern recording operation in the file system 12, when execution of the application 11 is started in the host 1, a data read request is sent from the process in the application 11 to the file management unit 121. When the file management unit 121 receives a data read request from the application 11 (step s51 in FIG. 5), it sends a notification to the access pattern recording unit 122 that the data has been read. Receiving this notification, the access pattern recording means 12 records information relating to data reading in the access pattern table stored in the access pattern storage unit 123 (step s52 in FIG. 5). Such an operation is repeated while the application 11 is being executed.

  Here, this access pattern recording operation will be specifically described with reference to an example until the access pattern table shown in FIG. 2 is completed.

  Since the host 1 performs parallel processing, when the execution of the application 11 is started, the process A and the process B are processed simultaneously. Process A and process B request the file management means 121 to read a file. First, process A requests reading of offset 0 to 4096 bytes of file a, and process B requests reading of offset 0 to 4096 bytes of file b. The file management unit 121 reads the access pattern recording unit 122 from the offset 0 to 4096 bytes of the file a, and the process B reads the offset 0 to 4096 bytes of the file b. To record that. The access pattern recording unit 122 creates a table for recording the access of the process A and the process B as an access pattern table, and file name = file a, offset = 0, size = 4096 with respect to the table for recording the access of the process A. And information such as file name = file b, offset = 0, size = 4096 is recorded in the table for recording the access of process B.

  Subsequently, process A requests reading of offset 4096 to 10240 bytes of file a, and process B requests reading of offset 4096 to 4096 bytes of file b. The file management unit 121 reads the access pattern recording unit 122 from the offset 4096 to 10240 bytes of the file a, and the process B reads the offset 4096 to 4096 bytes of the file b. To record that. The access pattern recording unit 122 records information that file name = file a, offset = 4096, size = 10240 in the process A table in the access pattern table, and file name = file b, in the process B table. Information of offset = 4096 and size = 4096 is recorded.

  Subsequently, process A requests reading of offset 14336 to 6144 bytes of file a, and process B requests reading of offset 8192 to 8192 bytes of file b. Similarly, the access pattern recording unit 122 records information that file name = file a, offset = 14336, size = 6144 in the process A table in the access pattern table, and file name in the process B table. = File b, offset = 8192, size = 8192 is recorded. In this way, an access pattern table as shown in FIG. 2 is completed.

  Next, FIG. 6 is a flowchart showing the operation of the data relocation unit 124 in the file system 12 of the present embodiment. As shown in FIG. 6, when the data rearrangement unit 124 in the file system 12 receives an instruction for data rearrangement sent from the application 11 when the application 11 ends (step s61 in FIG. 6), the access pattern The access pattern table is read from the storage unit 123 (step s62 in FIG. 6).

  Subsequently, the data reading order of the application 11 is specified based on the access pattern table (step s63 in FIG. 6), and the data stored in the disk device 2 is rearranged so as to be continuous in the specified order. (Step s64 in FIG. 6).

  For example, when a plurality of processes in the application 11 are executed at the same time and the access pattern storage unit 123 stores an access pattern table as shown in FIG. Based on the table, the total of the read size is calculated for each data read of the processes A and B, and the order of reading each data in the processes A and B is specified based on the calculated total, and in the specified order The data stored in the disk device 2 is rearranged so as to be continuous.

  When the access pattern storage unit 123 stores an access pattern table as shown in FIG. 4, the data rearrangement unit 124 is stored in the disk device 2 so as to be continuous in the order of the reading start time. Relocate data.

  Here, the application 11 instructs the file rearrangement unit 124 to start the file rearrangement. However, the file rearrangement unit 124 monitors the end of the application 11, and when the application 11 ends, The start of arrangement may be started.

  As described above, in the file system 12 of this embodiment, the data reading order of the application 11 is specified based on the access pattern table that is the data reading history of the application 11, and the data stored in the storage area of the disk device 2. Are sequentially rearranged in the order in which they are read by the application 11, even when a plurality of processes perform file access simultaneously when the application 11 is executed, the series of data access becomes sequential access, and the disk device 2 As a result, the application 11 can be speeded up.

  Here, in the configuration described above, since the access pattern table is created by the access pattern recording unit 122 when the application 11 is first executed, the file cannot be read at high speed when the application 11 is first executed. However, if the input / output pattern of the application 11 is known, an access pattern table created in advance may be stored in the access pattern storage unit 123. In this way, the data rearrangement unit 124 can perform data rearrangement based on the access pattern table before the application 11 is executed, and high-speed data access is performed from the first execution time of the application 11. be able to.

  Further, although all processes constituting the application 11 are configured to be executed on the host 1, the present invention is not limited to this, and the file system 12 is configured so that the process is distributed and executed on a plurality of hosts. It may be shared by a shared file system such as (Network File System). In this case, the access pattern recording unit 122 and the access pattern table are included in the NFS client, and the file relocation unit 124 is included in the NFS server.

  Further, the file management unit 121, the access pattern recording unit 122, and the data rearrangement unit 124 in the present embodiment may be configured such that the function contents are programmed and executed by a computer.

  As described above, according to the file system 12 of the present embodiment, data in the storage area of the disk device 2 is actually accessed based on the access pattern table that records the file access pattern when the application 11 is executed. Since the rearrangement is performed so that they are consecutive in the order in which they are performed, sequential access to the disk device 2 is performed when the application 11 is re-executed, and high-speed file access is possible.

  In the HPC field, the same input data file is used, and only the parameters are changed and the application is re-executed many times, and even when the input data file is changed and executed. In many cases, the access pattern does not change. Therefore, as in the file system 12 of the present embodiment, when the application 11 is executed for the first time, the access pattern of the input data file is recorded, and the data is rearranged according to the recorded access pattern. In subsequent executions, high-speed data access can be realized.

  In normal defragmentation technology, information in the storage device is arranged in a continuous area for each file, and it is not assumed that multiple processes perform file access at the same time. Doing so would cause the storage device to perform low-speed random access, resulting in slow file input / output. On the other hand, the file system 12 of the present embodiment rearranges the data stored in the disk device 2 so that they are consecutive in the order of reading by the application 11, so that each file is a fragment as shown in FIG. However, a series of file accesses performed by a plurality of processes when the application 11 is executed are sequential accesses in the disk device 2 and data input / output is performed at high speed. Thereby, the execution speed of the application can be effectively increased.

  The present invention can be applied to a computer system that executes parallel processing.

DESCRIPTION OF SYMBOLS 1 Host 2 Disk apparatus 11 Application 12 File system 121 File management means 122 Access pattern recording means 123 Access pattern storage part 124 Data rearrangement means 125 Data input / output part

Claims (12)

In a file system comprising file management means for executing read / write and deletion of a file stored in a disk device in response to a request from an application,
An access pattern storage unit for storing an access pattern table indicating a list of data read by the file management unit in response to a request from the application;
Data relocation means for relocating the data stored in the disk device based on the stored access pattern table so as to be continuously arranged in the order read by the application File system to be used. The file system according to claim 1, wherein
A file system comprising access pattern recording means for sequentially recording information relating to data read by the file management means in the access pattern table during execution of the application. The file system according to claim 2, wherein
The access pattern table is a table showing a reading file name, a reading start position, and a reading size in the order of reading for each process in the application,
The data relocation means, when a plurality of processes in the application are executed at the same time, based on the access pattern table, the function of calculating the total read size for each data read of each process, and this calculated It has a function of specifying the order of reading each data in all the processes based on each cumulative total, and a function of rearranging the data stored in the disk device so as to be continuously arranged in this specified order A file system characterized by that. The file system according to claim 2, wherein
The access pattern table is a table showing a read file name, a read start position, a read size, and a read start time for each process in the application in the order of reading.
The data rearrangement means has a function of specifying the order of reading each data in accordance with the start time of each data reading recorded in the access pattern table, and the disk so as to be continuously arranged in the specified order. A file system having a function of rearranging data stored in a device. A file management unit that executes reading / writing and deletion of a file stored in the disk device in response to a request from an application; an access pattern storage unit that stores an access pattern table indicating a list of data read by the file management unit; Is a file system with
The data relocation means reads the access pattern table from the access pattern storage unit,
Based on this access pattern table, the data relocation means identifies the data read by the application and the order in which it is read,
A data rearrangement method, wherein the data rearrangement means rearranges data stored in the disk device so as to be continuous in the specified order. The data rearrangement method according to claim 5, wherein
Before the data relocation means reads the access pattern table from the access pattern storage unit,
A data relocation method characterized in that, during execution of the application, the access pattern recording means sequentially records information relating to the data read by the file management means in the access pattern table. The data rearrangement method according to claim 6, wherein
The access pattern table is a table showing a read file name, a read start position, and a read size for each process in the application in the order of reading,
In specifying the order in which the data read at the time of executing the application is read,
When a plurality of processes in the application are executed at the same time, based on the access pattern table, the total read size is calculated for each data read of each process, and each process in all the processes is calculated based on each calculated total A data rearrangement method characterized by specifying an order of data reading. The data rearrangement method according to claim 7, wherein
The access pattern table is a table showing a read file name, a read start position, a read size, and a read start time for each process in the application in the order of reading,
When specifying the order of reading the data at the time of execution of the application,
A data rearrangement method characterized in that the order of reading each data is specified according to the start time of each data reading recorded in the access pattern table. A file management function for reading, writing, and deleting a file stored in the disk device in response to a request from an application;
Based on an access pattern table showing a list of data to be read by the file management function, a reading order specifying function for specifying data read by the application and its reading order; and
A data rearrangement program which causes a computer to execute a data rearrangement function for rearranging data stored in the disk device so as to be continuously arranged in the specified order. In the data rearrangement program according to claim 9,
A data rearrangement program for causing the computer to execute an access pattern recording function for sequentially recording information about data read into the file management function in the access pattern table during execution of the application. In the data rearrangement program according to claim 10,
The access pattern table is a table showing a read file name, a read start position, and a read size for each process in the application in the order of reading,
When a plurality of processes in the application are executed at the same time, the reading order specifying function calculates a total read size for each data read of each process based on the access pattern table. A data rearrangement program characterized by having a function of specifying the order of reading data in all the processes based on the above. In the data rearrangement program according to claim 11,
The access pattern table is a table showing a read file name, a read start position, a read size, and a read start time for each process in the application in the order of reading,
The read order specifying function is a function for specifying the order of reading each data in accordance with a start time of each data reading recorded in the access pattern table.

Images