JP4968218B2 - Disk storage device, data backup system, data relocation method, and program - Google Patents

Description

  The present invention relates to a disk storage device, a data backup system, a data relocation method, and a program, and more particularly to a disk storage device, a data backup system, a data relocation method, and a program for storing replication data obtained by replicating (replicating) backup target data. About.

  In recent years, as a data backup method for business storage, in order to reduce the business stop time during backup, backup from a replica created on a physical disk different from business data to a backup medium such as tape has been performed. .

  As the capacity of a physical disk increases, a method of creating a virtual physical disk called a logical LUN (Logical Unit Number) from one physical disk, setting a partition, and referring to the server is also generalized. For example, when the system is updated, it is possible to perform data migration without being aware of the difference in actual physical disk capacity by simply combining the logical LUN sizes.

  On the other hand, from the viewpoint of cost reduction, there is a tendency that only a small number of replica physical disks are mounted. Further, since RAID 5 (parity), which is cheaper than RAID 1 (mirror), tends to be adopted as the protection method, the number of logical LUNs cut out from the physical disk is several times the number of original (replication source) physical disks. .

  Patent Document 1 proposes an information processing system that, when writing data on a disk such as a magnetic disk or an optical disk, writes data in an empty sector closest to the last accessed sector, thereby reducing seek time. ing.

  In Patent Document 2, in a storage device that stores the same data in a plurality of rotary storage devices in a multiplexed manner, the data arrangement (inner A storage device is disclosed in which a control device executes access in accordance with the data arrangement when an access request is made.

Japanese Patent Laid-Open No. 10-283120 JP-A-9-134258

  FIG. 14 is a diagram showing the configuration of the disk storage device 100 in a business system having a business server A (11) and a business server B (12) used for business different from the business server A. Dashed lines 21, 22, and 23 represent physical disks, respectively. The physical disks 21 and 22 are configured with LUNs for the business server A (11) and other LUNs, and the physical disks 23 are configured with LUNs for the business server B (12) and other LUNs. In addition, the alternate long and short dash line 24 indicates that high speed is achieved by striping (RAID 0) using the LUN for the business server A (11) of the physical disks 21 and 22.

  Here, the physical disk 31 is a replica physical disk, and has a necessary and sufficient size capable of storing data of the three physical disks 21, 22, and 23. The backup server 40 creates a backup from the physical disk 31 to the tape device 41 according to a preset backup schedule.

  As illustrated in FIG. 14, when the number of logical LUNs included in the physical disks 21, 22, and 23 increases, a plurality of logical LUNs exist in the physical disk 31 that is the replication destination. In particular, as indicated by the alternate long and short dash line 24, when a plurality of logical LUNs are accelerated by the volume manager stripe (RAID0) function, the throughput is significantly reduced by simultaneous access to the plurality of logical LUNs. .

  FIG. 15 is a diagram showing the block arrangement status in the replication destination physical disk 31, and the numerical values in the figure show the access time for each block. As shown in FIG. 15, when replication is performed in units of physical disks, it is difficult to create a backup schedule in which no access occurs over a plurality of logical LUNs. For example, when a logical LUN for the business server A is backed up, access to the physical disk 31 is distributed over a wide range, and the seek time becomes long.

  The present invention has been made in view of the above-described circumstances, and the object of the present invention is to provide a disk storage device and data that can suppress a decrease in throughput when creating a backup from the replica physical disk. To provide a backup system, a data relocation method, and a program.

  According to the first aspect of the present invention, a disk medium that stores replication data obtained by replicating replication source data, and an access pattern to the disk medium are observed, and according to the data access order in the observed access pattern, There is provided a disk storage device comprising: means for determining a data arrangement order in the disk medium; and means for rearranging replication data according to the data arrangement order.

  According to the second aspect of the present invention, an access pattern to a disk medium storing replication data obtained by duplicating replication source data is observed, and the data in the disk medium is determined according to the data access order in the observed access pattern. There is provided a data rearrangement method for determining an arrangement order and rearranging the replication data according to the data arrangement order.

  According to the third aspect of the present invention, an access pattern to a disk medium that stores replication data obtained by duplicating replication source data is observed, and data in the disk medium is determined according to the data access order in the observed access pattern. There is provided a program for causing a computer to execute a process for determining an arrangement order and a process for rearranging the replication data according to the data arrangement order.

  According to the present invention, the cache hit rate increases and the throughput when reading data from the disk medium is improved. As a result, the time required for backup can be reduced. The reason is that focusing on the backup characteristics that backup is performed based on a predetermined backup schedule, the actual access order is set so that less seek time is required when the same access as the observed access status is performed. This is based on the localization of the data based on it.

[Summary of Invention]
First, the outline of the present invention will be described. FIG. 1 is a diagram schematically showing data rearrangement processing by the disk storage device according to the present invention. The diagram on the left side of FIG. 1 is a diagram showing the block arrangement status on the physical disk 31 before the relocation process, and the numerical values in the diagram show the access time (in units of 30 minutes) for each block (FIG. 15). The same).

  Therefore, the disk storage device according to the present invention observes access to the physical disk 31 and rearranges the physical disk data so that the seek time is shortened when the observed access pattern is reproduced. For example, the blocks accessed in the time zone of 21: 0 to 21:29 displayed as “21:00” in FIG. 15 are rearranged so as to be localized as shown on the right side of FIG. After the rearrangement, the data address after the rearrangement can be accessed from the data address before the rearrangement by using a pointer table or the like. As described above, the seek time for the next and subsequent accesses can be shortened. In addition, the cache hit rate can be improved.

[First Embodiment]
Next, a first embodiment of the present invention will be described in detail with reference to the drawings. FIG. 2 is a block diagram showing the configuration of one embodiment of the present invention. Referring to FIG. 2, a business server group 11/12, a backup server 40, and a connected disk storage device (disk storage device) 100 are shown.

  The disk storage device 100 includes a business magnetic disk group 21/22/23, a replica magnetic disk group 31A, a magnetic disk control mechanism 60, and a data rearrangement unit 61.

  The business magnetic disk group 21/22/23 is a disk group in which a logical LUN for the business server group 11/12 is set, and corresponds to the physical disks 21 to 23 in FIG.

  The replica magnetic disk group 31A is a disk group serving as a replication destination of the business magnetic disk group 21/22/23, and corresponds to the physical disk 31 of FIG.

  The magnetic disk control mechanism 60 responds to an access request from the business server group 11/12 and the backup server 40 to the business magnetic disk group 21/22/23 and the replica magnetic disk group 31A. When accessing the replica magnetic disk group 31A, a pointer table equivalent to that used for snapshots is referred to.

  The data rearrangement unit 61 observes access from the backup server 40 to the magnetic disk control mechanism 60 (access information), and creates a pointer table indicating the data arrangement order in the replica magnetic disk group 31A based on the result. ·Update. The data rearrangement unit 61 instructs the magnetic disk control mechanism 60 to rearrange data based on the created / updated pointer table. In this embodiment, based on the pointer table, data is rearranged by performing full replication from the business use magnetic disk group 21/22/23 to the physical disks included in the replica magnetic disk group 31A. .

  FIG. 3 is a block diagram illustrating an example of a specific configuration of the data rearrangement unit 61.

  The total block number storage unit 61A stores the number of blocks of the target physical disk included in the replica magnetic disk group 31A serving as a replication destination.

  The division number storage unit 61B stores a division number m of a physical disk that is a rearrangement processing unit to be described later. As the number of divisions of the physical disk, a value smaller than the number of blocks stored in the total block number storage unit 61A is set. This is because by adopting a rearrangement processing unit in which a certain number of blocks are collected, sorting and rearrangement processing described later are performed efficiently. Of course, the number m of physical disk divisions can also be set to the number of physical disk blocks.

  The block number storage unit 61C per area stores the number of blocks included in each area divided by the number of divisions of the physical disk. The number of blocks per area can be obtained by dividing the total number of blocks stored in the total block number storage unit 61A by the number of divisions stored in the division number storage unit 61B.

  The setting of the total number of blocks and the number of divisions may be omitted, and the user may directly set the number of blocks per area.

  As shown in FIG. 10, the first memory table 61D is configured by a table of m rows and n columns that records the number of accesses to each area for each time period. m is the number of divisions stored in the division number storage unit 61B, and n is determined by the observation time and the recording interval of the access time. For example, when the number of accesses is recorded in units of 15 minutes with an observation period of 24 hours, n = 24 [h] × 60 [minutes] ÷ 15 [minutes] = 96, and is divided into time zones t1 to t96. Is done.

  As shown in FIG. 11, the second memory table 61E refers to the first memory table 61D, and is an m-row / 2-column table that records the most frequently accessed time zone in each area.

  The first block number offset storage unit 61G, the second block number offset storage unit 61H, the area number storage unit 61I in the same time zone, the processing time zone storage unit 61J, and the block number counter 61K are various types used when creating the pointer table. Stores variables.

  The access monitoring means 62 is a means for monitoring access to the physical disk and recording it in the first memory table 61D.

  The access time zone extracting means 63 refers to the first memory table 61D and records the combination of the area number of each area and the most accessed time slot in each area in the second memory table.

  The pointer table creation means 64 includes a first block number offset storage unit 61G, a second block number offset storage unit 61H, a region number storage unit 61I in the same time zone, a processing time zone storage unit 61J, and a block in which processing variables are stored. The data arrangement order (pointer table) is created from the second memory table 61E with reference to the number counter 61K (see FIG. 13).

  The data rearrangement unit 61 and each processing unit described above can be configured by hardware, but can also be realized by a program executed by a computer installed in the disk storage device 100.

  Next, the flow of data rearrangement processing according to this embodiment will be described in detail with reference to the drawings.

[Advance preparation]
First, as an advance preparation, initial values are set and initialized as follows. FIG. 4 is an example of a process flowchart performed in advance preparation.

  First, the number of physical disk blocks is set in the total block number storage unit 61A (step J1).

  Next, the division number of the physical disk is set in the division number storage unit 61B (step J2).

  Next, the number of blocks included in one area is calculated and set in the block number storage unit 61C per area (step J3).

  Next, initialization processing is performed. Specifically, 0 is set in the first memory table 61D, the first block number offset storage unit 61G, and the area number storage unit 61I in the same time zone. -1 is set in the processing time zone storage unit 61J (step J4).

[Data relocation processing-sampling]
FIG. 5 is a flowchart showing the flow of the data rearrangement process. First, the data rearrangement unit 61 executes a sampling process for recording the number of accesses to each block to the physical disk and the time zone (step A1). The sampling process is performed by setting a certain sampling period (for example, 12 hours, 24 hours, etc.).

  FIG. 6 is a flowchart showing the flow of the sampling process. First, the data relocation unit 61 detects the block number and time of the access destination when the READ access has occurred to the target physical disk (step B1).

  Next, the access monitoring means 62 in the data rearrangement unit 61 adds 1 to the area number to which the detected block belongs and the value in the first memory table column corresponding to the time (step B2).

  The above processing is continued until the above-described sampling period elapses (No in step B3).

  FIG. 10 shows a first memory table in which access for each time period is stored in the sampling. In FIG. 10, t1 to tn indicate time zones set by predetermined sampling time intervals. For example, it can be read that the area number 3 receives 8 accesses in the time zone t3.

[Data relocation processing-Total access time]
Next, the access time zone extraction means 63 in the data rearrangement unit 61 is activated, refers to the first memory table 61D, searches for the time zone having the largest number of accesses for each area, and accesses the access time zone for that area. Is set (step A2 in FIG. 5).

  FIG. 7 is a flowchart showing the flow of access time zone extraction processing. First, the access time zone extracting means 63 reads one row from the first memory table 61D and extracts the time zone having the largest number of accesses (step C1).

  Next, the access time zone extracting means 63 stores the set of the read row area number and the extracted access time zone in the second memory table 61E (step C2). FIG. 11 shows a result of extracting a time zone in which the number of accesses is large for each row from the first memory table 61D of FIG.

  The above process is continued until the extraction of the access time period is completed for all the rows of the first memory table 61D (No in Step C3).

  When the extraction of the access time zone is completed for all the rows of the first memory table 61D (Yes in step C3), the pointer table creation means 64 is then activated.

  The pointer table creation means 64 sorts each row of the second memory table 61E in the order of the access time zone (step A3). FIG. 12 shows a state where the second memory table 61E of FIG. 11 is sorted (in ascending order) by the access time zone.

  The pointer table creation means 64 starts creating the data arrangement order (pointer table) using the sorted second memory table 61E (step A4).

  FIG. 8 is a flowchart showing the flow of processing for creating the data arrangement order (pointer table). First, the pointer table creation unit 64 reads one row from the second memory table 61E (step D1), and compares the access time zone of the row with the time zone stored in the processing time zone storage unit 61J (step S1). D2).

  Here, when the access time zone of the read row has changed (Yes in Step D2), the pointer table creation means 64 updates the value of the processing time zone storage unit 61J (Step D3) and the second The number of areas (number of rows) having the same access time zone is calculated with reference to the memory table 61E and stored in the area number storage unit 61I in the same time zone (step D4).

  Next, the pointer table creation means 64 sets the value of the second block number offset storage unit 61H to 0, and sets the value of the area number storage unit 61I in the same time zone to the first block number offset storage unit 61G (step S1). D5). The value in the second block number offset storage unit 61H is used for writing out a block number to be described later.

  If the access time zone of the read row has not changed (No in step D2), the pointer table creation unit 64 performs a process of adding 1 to the value of the second block number offset storage unit 61H (step D6). .

  When the above block number writing preparation is completed, the pointer table creating means 64 initializes the block number counter 61K with 0, and executes a process of writing the row read from the second memory table 61E in the data arrangement order (pointer table). (Step D7).

FIG. 9 is a flowchart showing a flow of processing for writing out the row read from the second memory table 61E in the data arrangement order (pointer table). First, the pointer table creating means 64 records the block number in the corresponding row of the data arrangement order (pointer table) 61F using the area number of the row read from the second memory table 61E as follows.
Row number: area number of read row × number of blocks per area + block number counter value Recorded content (block number): block number counter value × number of areas in the same time zone + first block number offset + second block number offset

  The number of blocks per area is stored in the block number storage unit 61C per area. The block number counter value is stored in the block number counter 61K. The number of areas in the same time zone is stored in the area number storage unit 61I in the same time zone. The first and second block number offsets are stored in the first block number offset storage unit 61G and the second block number offset storage unit 61H, respectively.

  The pointer table creating means 64 adds 1 to the value of the block number counter 61K (step E2). The above processing is continued until the value of the block number counter 61K reaches the number of blocks per area (step E3).

  When the writing of the block of the row read from the second memory table 61E is completed, the pointer table creating means 64 checks whether or not the writing of all the rows (all areas) of the second memory table 61E is completed (Step D8). ).

  Here, when writing of all the rows (all areas) of the second memory table 61E is not completed, the pointer table creating means 64 returns to step D1 and continues the processing for the next row of the second memory table 61E. (No in step D8).

  As a result, the block numbers included in each row are written according to the order in the second memory table 61E after sorting. FIG. 13 is an example of data arrangement information (pointer table) created from the second memory table 61E of FIG.

  Finally, when the creation / update of the data arrangement order (pointer table) is completed as described above, the data rearrangement unit 61 copies the data of all blocks from the business data physical disk to the replica physical disk. (Step A5). For example, as shown in FIG. 13, the seek time is minimized by rearranging the blocks of the physical disk 31, and high speed access is possible when the same access is performed from the next time.

  In particular, in this embodiment, a predetermined sampling period is set and the access status to the physical disk is observed, and from the result, the time zone with the highest number of accesses is extracted as the access time zone, and the data relocation plan is determined. Therefore, it is possible to perform highly effective data rearrangement.

  In the present embodiment, the data relocation is performed by copying the data of all blocks from the physical disk for business data to the physical disk for replica. However, the disk storage apparatus 100 is separately used for work. When a memory area is provided, data rearrangement may be performed using the area.

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments, and further modifications, replacements, and replacements may be made without departing from the basic technical idea of the present invention. Adjustments can be made. For example, the configuration and processing flowchart of the data rearrangement unit described above is merely an example, and it is possible to adopt a configuration and processing that can rearrange and localize data according to an actual access pattern. Is possible.

  For example, in the above-described embodiment, the access monitoring unit 62 is provided and the access pattern is obtained. However, the data access order may be derived based on the backup schedule obtained from the backup server 40 or the like. . Furthermore, if data rearrangement is performed at the timing when the backup schedule is updated, it is possible to suppress a decrease in throughput immediately after the backup schedule is updated.

  If the data storage device 100 of the above-described embodiment is linked to the backup server 40 shown in FIGS. 2 and 14, a data backup system capable of performing more efficient backup is provided.

It is the figure which represented typically the data rearrangement process by the disk storage apparatus which concerns on this invention. 1 is a block diagram showing the configuration of a disk storage device according to a first embodiment of the present invention. FIG. 3 is a block diagram illustrating an example of a specific configuration of a data rearrangement unit in FIG. 2. It is an example of the process flowchart performed in prior preparation. It is a flowchart showing the flow of the whole operation | movement of the 1st Embodiment of this invention. It is an example of the process flowchart performed in the sampling process of FIG. FIG. 6 is an example of a process flowchart performed in the access time zone extraction of FIG. FIG. 6 is an example of a process flowchart performed in creating the data arrangement order (pointer table) in FIG. 5. FIG. 9 is an example of a processing flowchart performed in recording of n block numbers in the area of FIG. 8. It is an example of a 1st memory table. It is an example of the 2nd memory table created from the 1st memory table. It is a figure showing the state which sorted the 2nd memory table of FIG. It is a figure showing an example of the pointer table. It is a figure showing the structure of the disk storage apparatus provided with the physical disk for replicas. It is a figure for demonstrating the mechanism which seek time increases.

Explanation of symbols

11 Business server A
12 Business server B
21, 22, 23 Physical disk (replication source)
24 RAID 0 (stripe) configuration 31 Physical disk (for replica)
31A Magnetic disk group (for replica)
40 Backup Server 41 Tape Device 60 Magnetic Disk Control Mechanism 61 Data Relocation Unit 61A Total Block Number Storage Unit 61B Division Number Storage Unit 61C Block Number Storage Unit 61D First Area Memory Table 61E Second Memory Table 61F Data Arrangement Order (Pointer table)
61G First block number offset storage unit 61H Second block number offset storage unit 61I Area number storage unit 61J Processing time zone storage unit 61K Block number counter 62 Access monitoring unit 63 Access time zone extraction unit 64 Pointer table creation Means 100 Disk storage device (disk storage device)

Claims (5)

A disk medium for storing replication data that is a copy of the original data;
Means for observing an access pattern to the disk medium and determining a data arrangement order in the disk medium according to a data access order in the observed access pattern;
Means for rearranging replication data according to the data arrangement order ,
The means for determining the data arrangement order in the disk medium is:
The storage area of the disk medium is divided by a predetermined size, and the number of accesses in a predetermined time segment for each divided area is recorded,
For each divided area, the time segment with the largest number of accesses is extracted as the access time zone of the divided area, and a table in which the divided areas are rearranged based on the access time zone is created. And
A disk storage device that refers to the table and creates a pointer table that sequentially assigns new block numbers to the blocks belonging to the areas having the same access time zone and indicates the correspondence between the block numbers before and after the rearrangement . 2. The disk storage device according to claim 1, wherein, when an access request to the disk medium is received, access is received with reference to the data arrangement order. The disk storage device according to claim 1 or 2 ,
A backup server that reads data from the disk storage device and performs data backup to a predetermined backup medium;
Including data backup system. By dividing the storage area of the disk medium storing the replication data obtained by copying the replication source data by a predetermined size, and recording the number of accesses in a predetermined time section for each of the divided areas , Observe access patterns,
For each divided area, the time segment with the largest number of accesses is extracted as the access time zone of the divided area, and a table in which the divided areas are rearranged based on the access time zone is created. And
By referring to the table, by assigning a new block number in order to the blocks belonging to the area where the access time zone matches, by creating a pointer table that represents the correspondence between the block numbers before and after the rearrangement, Determining a data arrangement order in the disk medium;
A data rearrangement method for rearranging the replication data according to the data arrangement order. By dividing the storage area of the disk medium storing the replication data obtained by copying the replication source data by a predetermined size, and recording the number of accesses in a predetermined time section for each of the divided areas , Processing to observe access patterns ;
For each divided area, the time segment with the largest number of accesses is extracted as the access time zone of the divided area, and a table in which the divided areas are rearranged based on the access time zone is created. Then, referring to the table, a new block number is assigned in order to the blocks belonging to the areas where the access time zones coincide with each other, and a pointer table representing the correspondence relationship between the block numbers before and after the rearrangement is created. Accordingly, the process of determining the data arrangement order in the disk medium,
A program that causes a computer to execute a process of rearranging the replication data in accordance with the data arrangement order.

Images