JP2003099208A - Method for data transfer between disk arrays and disk array system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To send a completion status back to a host in a processing time of the same level with normal processing to which mirroring is not applied. SOLUTION: Mirror transfer processing 204a by a mirror-source disk array 21 is carried out independently of host-side processing 202a and after the host- side processing 202a is completed, the completion status 205a is sent from the mirror-source disk array 21 to the host 11.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a first disk array system for storing data transferred from a host computer and a second disk array for storing a copy of data stored in the first disk array system. A mirror disk array system that includes a system and performs mirroring by transferring data transferred from a host computer and stored in a first disk array system from the first disk array system to a second disk array system. The present invention relates to a data transfer method suitable for use between disk arrays and a disk array system.

[0002]

2. Description of the Related Art Conventionally, two disk array systems have been used.
And one of the disk array systems (hereinafter referred to as the first disk array) stores data transferred from a host computer (hereinafter referred to as the host),
As the other disk array system (hereinafter referred to as the second disk array), a mirror disk array system configured to store a copy of data stored in the first disk array is known. Here, the host and the first and second disk arrays are SCSI (Small Comp
uter system interface) channel or SAN
Generally, they are interconnected by a network such as (Storage Area Network).

In the mirror disk array system, the host reads / writes data to / from the first disk array.
Write. The first disk array receives the data transferred from the host in response to the write request from the host and writes it to the cache memory in the own disk array. Disk write processing for storing data in (the disk of) its own disk array, and mirror transfer for transferring the data written in the cache memory to the second disk array and writing it to the same address of the second disk array Depending on the processing, mirroring that makes the contents of the first and second disk arrays the same is performed. Here, the first disk array is called a mirror source (mirror transfer source), and the second disk array is called a mirror destination (mirror transfer destination).

Usually, the first disk array sequentially executes the host side processing and the mirror transfer processing, and when the mirror transfer processing is completed and the data transfer to the second disk array is normally completed, When the write processing requested by the host is completed, a completion status is returned to the host. This ensures to the host that the contents of the first and second disk arrays are the same.

In this mirror disk array system,
Even if a failure occurs in the first disk array, the same data (mirror data) is stored in the second disk array, so the system operates normally by using the data in the second disk array. Can continue. That is, the mirror disk array system can improve the fault tolerance of the disk array.

[0006]

As described above, in the conventional mirror disk array system, the mirror source disk array sequentially executes the host side processing and the mirror transfer processing, and the mirror transfer processing is completed and the mirror destination is completed. When the data transfer to the disk array is completed normally, it is considered that the write processing requested by the host is completed and the completion status is returned to the host.

As described above, the method of returning the processing success status to the host due to the successful data transfer to the mirror destination disk array in response to the write request from the host to the mirror source disk array is reliable, but from the host. As seen, the processing time is almost twice as long as that of the normal disk array writing without mirroring, so that the processing performance is difficult. In particular, when the mirror transfer processing is performed from one mirror source disk array to a plurality of mirror destination disk arrays, the processing time becomes longer in proportion to the number of mirror destination disk arrays.

The present invention has been made in consideration of the above circumstances, and an object thereof is to carry out the mirror transfer processing in the mirror source disk array independently of the host side processing and to complete the host side processing. (EN) It is possible to provide a disk array data transfer method and a disk array system capable of returning a completion status to a host in the same level of processing time as normal processing that does not apply mirroring, by making a configuration that returns a completion status to the host. .

Another object of the present invention is to provide a disk array data transfer method and a disk array system capable of improving processing efficiency by switching processing procedures according to the progress of mirror transfer processing.

[0010]

According to the present invention, there is provided a first disk array (mirror source disk array) including a group of disk drives (disks) for storing data transferred from a host and the first disk array. What is claimed is: 1. A method of transferring data between disk arrays in a mirror disk array system comprising: a second disk array (mirror destination disk array) including a group of disks for storing a copy of data stored in the disk group; Executing a host-side process of receiving data transferred from the host to the first disk array in response to a write request and writing the data to the cache memory of the first disk array; Write the data written in the cache memory to the disk of the first disk array. And a step of executing a disk write process for loading and a mirror transfer process for performing mirroring by transferring the data written in the cache memory of the first disk array to the second disk array are independent of the host side process. And a step of returning a completion status for the write request from the first disk array to the host upon completion of the host side processing.

As described above, the mirror transfer process is asynchronous with the host process, and when the host process is completed, the completion status for the write request is returned to the host, thereby delaying the response to the host. As a result, when viewed from the host, the processing time for the write request can be set to the same level as the normal processing to which the mirroring is not applied.

Further, according to the present invention, the disk write processing is stopped according to the progress status of the mirror transfer processing, and by this disk write processing, the data which has been written to the disk of the first disk array and the mirror transfer processing has not been completed yet. Of the written area on the cache memory, releasing the area excluding some areas on the side where older data is written, and on the cache memory where data for which mirror transfer processing has not been completed is written When the data in the above-mentioned partial area becomes empty by the mirror transfer processing, the data written in the area on the released cache memory and already written in the disk of the first disk array is processed by the mirror transfer processing. To serve
And a step of reading from the disk of the first disk array into the partial area on the cache memory.

In such a configuration, there occurs a situation in which the processing speed of the mirror transfer processing cannot keep up with the processing speed of the host side processing, and as a result, the data of the mirror transfer processing incomplete in the cache memory increases. Also, since the area in the cache memory where the data transferred from the host can be written, that is, the area in the cache memory that is the target of writing in the host side processing can be expanded, the cache memory usage efficiency can be improved and the host The processing time can be shortened when viewed from above. Here, the processing amount increases because it is necessary to read the data once written to the disk of the first disk array (mirror source disk array) into the cache memory again, but the processing speed of the mirror transfer processing is Even if a situation where the processing speed of the processing cannot keep up, the contents of the first disk array and the second disk array are quickly matched, so the processing of the second disk array (mirror destination disk array) is performed. Due to factors such as capacity, load status, and usage rate of the transmission path serving as a mirror path, even if a delay or an error occurs, the probability of causing a problem is low, and as a result, processing efficiency can be improved.

Further, according to the present invention, when the first disk array is stopped, management information regarding data which has been written to the disk of the first disk array and for which the mirror transfer processing has not been completed is secured in the disk. In the existing management information area, and when the first disk array is restarted, corresponding data is transferred from the disk of the first disk array to the cache memory based on the management information saved in the management information area. And a step of restarting the mirror transfer processing for the data read in the cache memory.

In such a configuration, even if a situation occurs in which the first disk array is stopped, when the first disk array is restarted, the mirror transfer process is restarted / continued and the first disk array is operated with a minimum load. It is possible to match the contents of the first disk array and the contents of the second disk array. Here, if the size of the management information exceeds a certain size, the corresponding data amount increases, but the size of the management information becomes smaller than the certain size. If switching to the use of new management information, the management information is saved. Since the required time can always be kept within a certain time, it is possible to increase the probability that the management information can be surely saved when the situation where the first disk array is stopped occurs.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a block configuration of a computer system including a mirror disk array system according to an embodiment of the present invention.

In FIG. 1, hosts (host computers) 11, 12, 13 for processing data and two disk arrays (disk array systems) 21, 22 constituting a mirror disk array system are a network 30 such as a SAN. Connected by. The disk array 21 is a mirror source disk array that stores data processed by the hosts 11 to 13.
Reference numeral 2 is a mirror destination disk array for storing a copy of data stored in the disk array 21.

The disk array 21 is, for example, two disk drives (hereinafter referred to as disks) 211-1 and 2-2.
11-2 and a controller 213 that reads / writes data from / to the disks 211-1 and 211-2 in response to a read / write request (read / write command) from the hosts 11 to 13. The disks 211-1 and 211-2 are invisible to the user of the disk array 21, that is, the host 1 is instructed by the user.
System management information area 2 that cannot be accessed from 1 to 13
12-1 and 212-2 are secured. Controller 21
3 is a CPU 21 which is the center of the controller 213.
4 and R that secures the work area and the like of the CPU 214.
It is composed of an AM 215 and a cache memory 216. The cache memory 216 is used by the hosts 11 to 13.
It is used to temporarily hold the data transferred from the hosts 11 to 13 in response to a write request from the disk 211-i (i is 1 or 2).

The disk array 22 is the disk array 21.
Similarly to the above, it is composed of the disks 221-1 and 221-2 and the controller 223. Controller 223 is CP
It is composed of a U 224, a RAM 225, and a cache memory 226.

Next, regarding the operation in the system of FIG. 1, a write request from the host 11 is issued by the disk array 2
1 will be described as an example with reference to FIG. 2 shows the host 11 to the mirror source disk array 21.
3 is a sequence chart for explaining an operation procedure when a write request is given to the conventional method in comparison with a conventional method.

The hosts 11 to 13, for example the host 11,
Data is read / written via the network 30 by issuing a read / write request to the disk array 21.

Disk array 21 (within controller 2
13) is a write request (command) 201a from the host 11 as shown in the sequence chart of FIG.
Is given to the network 3 from the host 11.
Host-side process 202a that receives data transferred via 0 and writes it in the cache memory 216 in the own disk array 21, and the data written in the cache memory 216 by this host-side process 202a in the disk in the own disk array 21. The disk write processing 203a for writing to the disk 211-i and the data written in the cache memory 216 are transferred to the disk array 22 via the network 30 and the disk 2 in the disk array 22 concerned is transferred.
Mirror transfer processing 2 to write to the same address of 21-i
04a and the disk 21 in the disk array 21.
Mirroring is performed so that the contents of 1-i and the disks 221-i in the disk array 22 are the same.

Here, the disk array 21 (the controller 213 therein) executes the mirror transfer processing 204a independently (that is, asynchronously) with respect to the host-side processing 202a, not sequentially. And disk array 2
As shown in FIG. 2A, the controller 1 (internal controller 213) returns the completion status 205a to the host 11 regardless of the progress of the mirror transfer processing 204a when the host-side processing 202a is completed. Incidentally, in the conventional case, the disk array 21 sequentially performs the mirror transfer process 204b with respect to the host side process 201b as shown in FIG. 2B, and when the mirror transfer process 204b is completed, that is, the disk array. 22 to disk array 2
When the completion status 206b was returned to 1, the completion status 205b was returned to the host 11.

Therefore, in this embodiment, the write request given from the host 11 to the disk array 21 is viewed from the host 11 and the processing time is about 1/2 of that of the conventional method shown in FIG. In other words, the processing can be performed in the same processing time as the normal disk array writing without mirroring.

In the mirror transfer processing, a delay or an error may occur due to factors such as the processing capacity and load status of the mirror destination disk array 22 and the usage rate of the transmission path (network 30) which is the mirror path. In such a case, if the mirror transfer process is executed independently of the host side process as in the present embodiment,
Even when the completion status (normal completion status) is returned to the host upon completion of the processing on the host side, there is a case where the mirror destination disk array 22 does not reflect the normal completion. Therefore, in order to maintain consistency between the host-side processing and the mirror transfer processing, for example, data in the entire area of the disk 211-i in the mirror source disk array 21 must be transferred to the mirror destination disk array 22. As a result, processing efficiency is reduced. Therefore, in the present embodiment, the processing procedure in the mirror source disk array 21 including the mirror transfer processing is switched to three stages of processing A, B, and C according to the progress status of the mirror transfer processing, as described below. The processing efficiency of the mirror source disk array is improved.

First, the process A will be described with reference to the operation explanatory diagram of FIG. 3 and the flowchart of FIG. The flowchart of FIG. 8 shows the procedure of process A. Host 11
When a write request is given from the mirror source disk array 21 to the mirror source disk array 21,
The CPU 214 provided in 13 writes the write data to the cache memory 216 every time the write data transferred from the host 11 is received (steps S1 and S).
2). When the disk array 21 normally receives all the write data specified by the write request and completes the writing to the cache memory 216, that is, when the host side processing is completed, the completion status for the write request is sent to the host 11. (Normal end status) is returned (steps S3 and S4).

The mirror source disk array 21 is the host 1
When the completion status is returned to 1, the disk write processing (step S5) of writing the data written in the cache memory 216 to the disk 211-i (i is 1 or 2) in the own disk array 21 and the cache memory 216 The data written in the disk array 22 is transferred to the mirror destination disk array 22 and
The mirror transfer processing (step S6) for writing the same address 21-i is started. The disk write process (step S5) and the mirror transfer process (step S5)
6) at the time before the completion status notification to the host 11,
For example, the configuration may be started when the first data designated by the write request from the host 11 is written in the cache memory 216.

When the mirror source disk array 21 returns the completion status to the host 11, it becomes ready to accept the next request from the host 11. In this state,
When a new write request is sent from the host 11 to the disk array 21, the disk array 21 receives the request, receives write data newly transferred from the host 11 and writes the write data in the cache memory 216 (step S1). , S2) is executed again. As a result, in the mirror source disk array 21, the host side processing, the disk write processing and the mirror transfer processing are performed in parallel.

In such a state, the processing speed of the mirror transfer processing, that is, the processing speed of the data transfer from the mirror source disk array 21 to the mirror destination disk array 22 is the host side processing speed, that is, the host 11 to the mirror source disk array. A situation may occur in which the processing speed of data transfer to the T.21 cannot be kept up. In that case, the amount of mirror-transfer-uncompleted data on the cache memory 216 increases, and the utilization efficiency of the cache memory 216 decreases.

Therefore, in this embodiment, of the data on the cache memory 216, the mirror source disk array 21
Total amount of data (total size) that has been written to disk 211-i inside (disk written) and mirror transfer has not completed
When S exceeds a predetermined fixed size Sa, the process B is switched to the process B including a procedure that enables improvement of utilization efficiency of the cache memory 216.

Therefore, the mirror source disk array 21 is
For example, every time a certain amount of data written in the cache memory 216 is written to the disk 211-i in the own disk array 21 by the disk write process (step S
5) The size S of the data in the cache memory 216 which has been written to (the disk 211-i of) the disk array 21 and mirror transfer to the mirror destination disk array 22 has not been completed is checked (step S7). Here, of the data in the cache memory 216, a queue of data that has been written to the disk 211-i in the mirror source disk array 21 and has not been mirror-transferred is set as the transmission queue 31.
I will call it.

In the mirror source disk array 21, the size S, that is, the size S of the current transmission queue 31 does not exceed the predetermined fixed size Sa (step S
8) and (the disk 211 of its own disk array 21
If there is data to be written in -i) (step S9), the disk write process (step S5) is continued.

On the other hand, the size (transmission queue 31
Size) S, that is, if the size S of the data that has been written to the disk and mirror transfer has not been completed exceeds a certain size Sa (step S8), the mirror source disk array 21 stops the disk write process (step S5) and will be described later. The process is switched to the process B (step S11). Further, the mirror source disk array 21 transfers the mirror-finished data (data in the transmission queue 31) on the cache memory 216 to the mirror destination disk array 22 regardless of switching to the process B (step S6). ) Is continued (step S10).

Next, the process B will be described with reference to the operation explanatory diagram of FIG. 4 and the flowchart of FIG. The flowchart of FIG. 9 shows the procedure of process B. When the mirror source disk array 21 (the CPU 214 provided in the controller 213 therein) is switched to the process B, the older one of the areas of the size S on the cache memory 216 (the area of the current transmission queue 31) is selected. Fixed size Sb on the data storage side (however, Sb <Sa <S)
The area excluding the area, that is, the area of size S-Sb is released (step S21). As a result, the transmission queue 31
While the size of S is reduced from S to Sb, the area in which the data transferred from the host can be written in the cache memory 216 is expanded by the size S−Sb, and the utilization efficiency of the cache memory 216 is improved.

When the area of size S-Sb on the cache memory 216 is released, the mirror source disk array 21 stores the storage destination disk address (first disk address) of the mirror-transfer-uncompleted data written in that area. The size is registered in the entry of the update list 41 (step S22). The area of the update list 41 is R provided in the controller 213 of the disk array 21.
Reserved on AM215. FIG. 6 shows an example of the data structure of the update list 41. The area of the update list 41 may be used by switching to the area of the update table 51j as described later.

The disk array 21 determines whether or not the size of the update list 41 exceeds a predetermined fixed size as a result of registration of the entry information in the update list 41 (step S23). The reason for performing this determination step S23 will be described later.

If the size of the update list 41 does not exceed the fixed size, the mirror source disk array 21
Is an area of the size Sb on the cache memory 216,
That is, it waits until the area of the new transmission queue 31 becomes empty by the mirror transfer processing (step S6) (step S24). Then, in the mirror source disk array 21, as a result of all the data in the area of the size Sb (area of the new transmission queue 31) on the cache memory 216 being transferred to the mirror destination disk array 22, the area of the size Sb is in an empty state. If it becomes, the disk 211 in the own disk array 21 is read according to the entry information of the update list 41.
-Read the corresponding size S-Sb data from i,
The data is written in the area of the size Sb on the cache memory 216 (step S25). Normally, S-Sb is smaller than Sb, so that all data of size S-Sb can be written in the area of size Sb. Note that S-
When Sb is larger than Sb, the data of size Sb of the data of size S-Sb is written in the area of size Sb, and the data of size Sb is subjected to the mirror transfer process (step S6). Then, the remaining data may be written in the area of size Sb.

The data designated by the entry information of the update list 41 is read from the disk 211-i in the mirror source disk array 21 and passed through the area of the size Sb (area of the new transmission queue 31). When all data is transferred to the mirror destination disk array 22 (step S2
6), the mirror source disk array 21 deletes the entry information from the update list 41 (step S27).
The mirror source disk array 21 is subjected to the above step S25.
The process from S27 to S27 is repeated until the update list 41 becomes empty (step S28).

In this way, with the disk write processing (step S5) stopped, only the area of the size Sb is used as the area of the transmission queue 31, and the disk array 2
The data that has been written to the disk 211-i in 1 and the mirror transfer is not completed is transferred from the disk 211-i through the area of the size Sb to the mirror destination disk array 22.
The processing B for transferring to the cache memory 216 can improve the utilization efficiency of the cache memory 216. However, in the process B, the disk 2 of the mirror source disk array 21 is once
Since a process (step S25) for re-reading the data written in 11-i is required, the processing amount is increased as compared with the process A and the processing efficiency in the disk array 21 is reduced.

Therefore, when the update list 41 becomes empty (step S28), the mirror source disk array 21
That is, when all the data that has been written to the disks 211-i in the own disk array 21 and the mirror transfer has not been completed are transferred to the mirror destination disk array 22, the process B
Then, the process returns to the process A (step S29).

Next, a case where the size of the update list 41 exceeds a predetermined fixed size will be described. The size of the update list 41 is the RAM for the update list 41.
When the size (a fixed size) of the area secured on the 225 is exceeded, the entry information (list data) forming the update list 41 cannot be stored in the RAM 225. In this case, the process B cannot be continued. Therefore, when the mirror source disk array 21 determines that the size of the update list 41 exceeds a certain size (step S2
3) The process is switched to the process C in which the size of the area on the RAM 225 required for the operation is smaller than that in the process B.

Next, the process C will be described with reference to the operation explanatory diagram of FIG. 5 and the flowchart of FIG. In addition,
The flowchart of FIG. 10 shows the procedure of process C. When switching to the process C, the mirror source disk array 21 (the CPU 214 provided in the controller 213) updates the table 51j (j) based on the current update list 41.
= P, q, r ...) is created (step S31). Figure 7
Shows the relationship between the area of the disk 211-i and the update table.

In process C, the area of the disk 211-i is managed by dividing it into blocks of the first size (large blocks having a small resolution). Further, the block of the first size (large block) is divided into units of the block of the second size (small block having a large resolution) and managed.
The second size is 1 / n of the first size (n is an integer of 2 or more, for example, 8). That is, the area of the disk 211-i is managed in units of blocks that are hierarchical (here, hierarchized in two stages).

The update table 51p is the disk 211-i.
Each of the large blocks constituting the area is composed of bits, and the presence / absence of update for each large block is represented by an ON / OFF state (1/0) of the corresponding bit. The update tables 51q, 51r ... Are prepared for each large block forming the area of the disk 211-i, and are composed of bits respectively corresponding to n small blocks forming the large block, and update for each small block. The presence or absence of is indicated by the ON / OFF state of the corresponding bit.

As described above, in the update table applied in this embodiment, the update table 51p and the update table 51p are provided corresponding to the blocks (the large block and the small block) hierarchized in two stages.
There are two types of update tables 51q, 51r, ... For a large number of blocks. That is, in this embodiment, a multi-stage table configuration in which the resolution of blocking is changed is applied to the configuration of the update table.

The update table 51j (j = p, q, r ...) Created at the start of the process C is processed by the process C as described later.
In addition to being operated in the middle (OFF operation), it is also operated (ON operation) in the host side processing. That is, when the mirror source disk array 21 receives a write request from the host,
For the block to which the disk area specified by the write request belongs or all the blocks included in the disk area, the corresponding bit of the update table 51j (j = p, q, r ...) Is turned ON. Note that this bit operation may be performed at the time of executing the disc write process (step S5).

When the update table 51j is created based on the update list 41, that is, when the update list 41 is converted into the update table 51j, the mirror source disk array 21 refers to the update table 51j and displays the update table 51j. In step S32, the update block is searched from the bit in the ON state. Then, if the update block can be retrieved (step S33), the mirror source disk array 21 determines that the disks 21 in the own disk array 21 are the same.
The data of the update block is read from 1-i and written in the area of the size Sb (area of the transmission queue 31) on the cache memory 216 (step S34). The updated block data written in the area of size Sb (area of the transmission queue 31) is transferred to the mirror destination disk array 22 in the mirror transfer process (step S6).

When the mirror transfer of the update block is completed (step S35), the mirror source disk array 21 turns off the corresponding bit in the update table 51j (step S36), and then executes step S32 again, and then Search for the updated block that is the target of the mirror transfer of.

The mirror source disk array 21 performs the above steps S32 to S36 until all the update blocks to be mirror-transferred cannot be searched, that is, all the bits in the ON state on the update table 51j are OFF.
This is repeated until the status is reached and it is confirmed that the mirror transfer of all update blocks is completed.

As described above, only the area of the size Sb is used as the area of the transmission queue 31, and the update table 51j is used.
The data managed as an update block by the mirror source disk array 21 is transferred from the disk 211-i in the mirror source disk array 21 via the area of the size Sb.
By the process C of transferring the data to No. 2, the utilization efficiency of the cache memory 216 can be improved while using the update table 51j having a size smaller than the update list 41.

However, in the process C, the data once written to the disk 211-i of the mirror source disk array 21, the data of the block that has not been actually updated, or the block including the data that has not actually been updated. Processing for reading data (step S34)
Therefore, the processing amount increases as compared with the processing A and B, and the processing efficiency in the disk array 21 decreases.

Then, the mirror source disk array 21 returns from the process C to the process A when the update blocks to be the mirror transfer cannot be retrieved as a result of the completion of the mirror transfer of all the update blocks (step S33). Yes (step S37). As a result, the process (disk write process) of step S5 in the process A is restarted, and the data in the cache memory 216 not written in the disk 211-i in the mirror source disk array 21 is stored in the disk 211-i. Written.

By the way, the update table 5 having a small resolution
With 1p, the size of the table itself can be small, but since the block size is large, the size of mirror transfer is also large when an update occurs, resulting in poor efficiency. Therefore, the processing efficiency of the mirror transfer is higher when the update tables 51q, 51r, ... Having a higher resolution are used.

On the other hand, when the mirror source disk array 21 is stopped due to a power failure or the like, the update table 5 is used for processing when the disk array 21 is restarted, as will be described later.
When attempting to save 1j in the system management information area 212-i, there may be a situation in which not all update tables 51j can be saved. In this case, the update table 51j having a small size, that is, the update table 51p having a small resolution is used.
If is preferentially stored, the update table 51p can be reliably stored. As a result, when the disk array 21 is restarted, the update table 5 stored in the system management information area 212-i is updated.
By using 1p, although the mirror transfer processing efficiency is not sufficient, it becomes possible to restart / continue the mirror transfer processing.

Therefore, in this embodiment, the update table is
As a multi-stage table configuration of an update table 51p having a low resolution and update tables 51q, 51r, ... Having a high resolution, the blocks are hierarchically managed into a large block and a small block, and the above-mentioned steps in operation of the mirror source disk array 21 are performed. In the process of S32, the update block is searched using the update tables 51q, 51r, ... Having a large resolution in consideration of the mirror transfer efficiency. Further, when the mirror source disk array 21 is stopped, the update table 5 having a small resolution is used.
1p is the system management information area 212 of the disk 211-i.
By saving preferentially to -i, the saving of the update table as management information is ensured.

Next, the operation in the case where a situation in which the disk array 21 is stopped, such as the power-off of the mirror source disk array 21, has occurred will be described with reference to the flowchart of FIG. The flowchart of FIG. 11 shows a processing procedure when the mirror source disk array 21 is stopped.

When the mirror source disk array 21 is stopped, if the process A is being executed (step S41), the transmission queue 31 on the cache memory 216 is transmitted.
Information on the data of the disk, that is, own disk array
The disk address and the size of the data which has been written to the disk 211-i in the internal disk array 21 and which has not been subjected to the mirror transfer are used as the system management information (mirror transfer incomplete area information) to be used at the time of restarting itself.
The data is saved in the system management information area 212-i of 1-i (step S42). If the process B is being executed (step S43), the mirror-source disk array 21 has a RAM.
The update list 41 stored in 215 is stored in the system management information area 212-i of the disk 211-i as the system management information (mirror transfer incomplete area information) (step S44). In addition, the mirror source disk array 21
If neither process A nor process B is executed,
That is, if the process C is being executed (step S41,
S43), the update tables 51p, 51q, 51r, ... Stored in the RAM 215 are given the highest priority, and the system management of the disk 211-i is performed as the system management information (mirror transfer incomplete area information). The data is stored in the information area 212-i (step S45).

Next, the operation of restarting the mirror source disk array 21 once stopped will be described with reference to the flowchart of FIG. The flowchart of FIG. 12 shows a processing procedure when the mirror source disk array 21 is restarted.

Now, it is assumed that the mirror source disk array 21 has been restarted. In this case, the mirror source disk array 21 refers to the system management information area 212-i of the disk 211-i in the own disk array 21, and refers to the system management information area 212-i (system transfer information is not completed). It is checked whether the area information) is correctly stored (steps S51 and S52).

If the system management information is correctly stored, the mirror source disk array 21 reads the corresponding data from the disk 211-i in the own disk array 21 based on the system management information and caches it. Mirror destination disk array 2 via memory 216
The mirror transfer process for transferring the data to No. 2 is restarted / continued (step S53).

If the update table is stored as the system management information, the mirror source disk array 2
First, it checks whether or not the update tables 51q, 51r, ... With large resolution that enable efficient mirror transfer processing are stored, and if so, based on the update tables 51q, 51r ,. Performs mirror transfer processing. Further, when the update tables 51q, 51r, ... Are not stored, the mirror source disk array 21 checks whether or not the update table 51p having a small resolution is stored, and if it is stored, based on the table 51p. Performs mirror transfer processing. As described above, when the mirror source disk array 21 stops during the execution of the process C, the saving is shorter than the update table 51p having a small resolution, that is, the update tables 51q, 51r, ... Having a large resolution. The update table 51p that can be changed over time is saved with priority. Therefore, in the system management information area 212-i,
Even if the update tables 51q, 51r, ... Are not correctly saved, the update table 51p is highly likely to be correctly saved. In this case, the update table 51 stored in the system management information area 212-i
Although the transfer efficiency is not good based on p, it is possible to perform the mirror transfer processing. Therefore, the contents of the disks 211-i in the mirror source disk array 21 and the disks 221-i in the mirror destination disk array 22 are Can be matched.

On the other hand, when the system management information is not correctly saved, that is, when the saving of the system management information when the mirror source disk array 21 is stopped has failed (step S52), the disk array 21: Although the mirror transfer efficiency is extremely poor, the own disk array 21
The data of the entire area of the internal disk 211-i is copied to the mirror destination disk array 22 and an attempt is made to restore (step S54).

By the way, as an example of failure in saving the system management information (mirror transfer incomplete area information) when the mirror source disk array 21 is stopped, the size of the system management information is one recording unit of the disk 211-i ( sector)
Because of the larger size, the mirror source disk array 21 may be completely stopped, for example, at the stage of saving one recording unit, that is, in the state where the system management information is saved halfway. In this case, if the size of the system management information is one recording unit or less, it means that the system management information can be saved before the mirror source disk array 21 is completely stopped.

Therefore, instead of the condition applied in the above embodiment, that is, the condition that the size S of the transmission queue 31 exceeds the size Sa, as the condition for switching from the process A to the process B, the management information of the data of the transmission queue 31 is changed. The condition that the size exceeds one recording unit may be applied. In this case, when the situation where the mirror source disk array 21 stops during the execution of the process A, the size of the management information regarding the data of the transmission queue 31 to be saved in the system management information area 212-i is recorded as one. It can be kept below the unit.

Further, as a condition for switching from the process B to the process C, the fixed size under the condition applied in the above embodiment, that is, the size of the update list 41 exceeds the fixed size may be set as one recording unit. in this case,
The size of the update list 41, which is the storage target information during execution of the process B, can be suppressed to one recording unit or less.

The size of the large block may be determined so that the size of the update table 51p, which is the storage target information during execution of the process C, is one recording unit or less.

In the above embodiment, the host 11 (1
2, 13) and the two disk arrays 21 and 22 constituting the mirror disk array system are the network 30.
However, the present invention is not limited to this. For example, as shown in FIG. 13, even if the host 11 and the disk arrays 21 and 22 are connected by channels 131a and 131b such as SCSI, the present invention can be similarly applied.

The present invention is not limited to the above-described embodiment, and can be variously modified in an implementation stage without departing from the spirit of the invention. Furthermore, the embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some of the constituent elements are deleted from all the constituent elements shown in the embodiment, at least one of the problems described in the section of the problem to be solved by the invention can be solved and described in the section of the effect of the invention. When at least one of the effects described above is obtained, a configuration in which this constituent element is deleted can be extracted as an invention.

[0069]

As described in detail above, according to the present invention, the mirror transfer processing in the mirror source disk array is processed independently of the host side processing, and is completed to the host when the host side processing is completed. With the configuration that returns the status, the delay of the response to the host can be suppressed, and the completion status can be returned to the host in the same processing time as the normal processing that does not apply mirroring.

Further, according to the present invention, the processing efficiency can be improved by switching the processing procedure according to the progress status of the mirror transfer processing, and the processing speed of the mirror transfer processing can catch up with the processing speed of the host side processing. Even if a situation does not occur, the utilization efficiency of the cache memory can be improved, and the processing time can be shortened from the viewpoint of the host.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a computer system including a mirror disk array system according to an embodiment of the present invention.

FIG. 2 is a sequence chart for explaining an operation procedure when a write request is given from the host 11 to the mirror source disk array 21 in the embodiment, in comparison with a conventional method.

FIG. 3 is a view for explaining the operation of process A in the same embodiment.

FIG. 4 is a view for explaining the operation of process B in the same embodiment.

FIG. 5 is a view for explaining the operation of process C in the same embodiment.

FIG. 6 is a diagram showing an example of a data structure of an update list 41 used in a process B.

FIG. 7 is a diagram for explaining the relationship between the area of the disk 211-i and the update table 51j (j = p, q, r ...) Used in the process C.

FIG. 8 is a flowchart showing a procedure of processing A.

FIG. 9 is a flowchart showing the procedure of processing B.

FIG. 10 is a flowchart showing the procedure of processing C.

FIG. 11 is a flowchart showing a processing procedure when the mirror source disk array 21 is stopped.

FIG. 12 is a flowchart showing a processing procedure when the mirror source disk array 21 is restarted.

FIG. 13 is a diagram showing a modification of the system shown in FIG.

[Explanation of symbols]

11, 12, 13 ... Host (host computer) 21 ... Disk array (mirror source disk array system) 22 ... Disk array (mirror destination disk array system) 30 ... Network 41 ... Update list 51p, 51q, 51r, 51j ... Update table 131a, 131b ... Channels 211-1, 211-2, 221-1, 221-2, 211-i ...
Disk (disk drive) 213, 223 ... Controller 216, 226 ... Cache memory

Claims (18)

[Claims] 1. A first disk array system including a group of disk drives for storing data transferred from a host computer and a disk for storing a copy of data stored in the disk drive group of the first disk array system. A method for transferring data between disk arrays in a mirror disk array system including a second disk array system including a group of drives, the host disk including the first disk array in response to a write request from the host computer. Performing a host-side process of receiving data transferred to the system and writing the data to the cache memory of the first disk array system; and writing the data written to the cache memory of the first disk array system to the first disk array system. Disc array A step of executing a disk write process for writing to the disk drive of the system; and a mirror transfer process for performing mirroring by transferring the data written in the cache memory of the first disk array system to the second disk array system. And a step of returning the completion status for the write request from the first disk array system to the host computer when the host side processing is completed. A method for transferring data between disk arrays, characterized by: 2. A step of detecting the progress status of the mirror transfer processing, the disk write processing being stopped according to the progress status of the mirror transfer processing, and the disk write processing of the first disk array system. A step of releasing an area of the area on the cache memory in which the data that has been written to the disk drive and the mirror transfer processing has not been completed is removed, except for a partial area on the side where older data is written; If the data in the partial area on the cache memory in which the data for which the mirror transfer processing has not been completed is written becomes empty by the mirror transfer processing, it is written in the released area on the cache memory. The disk drive that has already been written to the disk drive of the first disk array system. Data is read from the disk drive of the first disk array system into the partial area of the cache memory in order to be used for the mirror transfer processing. Data transfer method between disk arrays. 3. In the detecting step, from the total size S of data in the cache memory that has been written to the disk drive of the first disk array system by the disk write processing and the mirror transfer processing is incomplete, The progress of the mirror transfer process is detected, and in the releasing step, when the total size S exceeds a predetermined constant size Sa, the disk write process is stopped to release the area. The data transfer method between disk arrays according to claim 2. 4. The size of the partial area is a predetermined fixed size Sb, and in the releasing step, the size Sb of the areas of the size S on the cache memory of the first disk array system is used. 4. The method of transferring data between disk arrays according to claim 3, wherein the area of size S-Sb excluding a part of the area is released. 5. The method further comprises a step of registering entry information including a disk address and a size of data of the released area in an update list, wherein in the reading step, the partial area on the cache memory is the mirror. The data specified by the entry information of the update list is read from the disk drive of the first disk array system to the partial area of the cache memory when the transfer processing becomes empty. The data transfer method between disk arrays according to claim 2. 6. When the data specified by the entry information of the update list is transferred from the partial area on the cache memory to the second disk array system by the mirror transfer processing, the entry information is transferred to the second disk array system. 6. The method of transferring data between disk arrays according to claim 5, further comprising a step of deleting from the update list and a step of restarting the disk write process when the update list becomes empty. 7. When the size of the update list exceeds a predetermined fixed size, the area of the disk drive of the first disk array system is divided into blocks of a predetermined size from the update list. For updating and managing for each block, a step of creating an update table of a size smaller than the certain size, a step of searching for an update block according to the update table, and a step of searching the update block In this case, the method further comprises the step of reading the data of the updated block from the disk drive of the first disk array system into the partial area of the cache memory in order to use it for the mirror transfer processing. 6. The method of transferring data between disk arrays according to claim 5. 8. When the data of the update block searched according to the update table is transferred from the partial area on the cache memory to the second disk array system by the mirror transfer processing, the update table is updated. 8. The method according to claim 7, further comprising: a step of operating to change the corresponding block to indicate non-update, and a step of restarting the disk write process when the updated block cannot be searched. Data transfer method between disk arrays. 9. The method of transferring data between disk arrays according to claim 7, wherein the fixed size is a size of a memory area reserved for storing the update list or the update table. 10. The progress of the mirror transfer process regarding data that has been written to the disk drive of the first disk array system and the mirror transfer process has not been completed when the first disk array system is stopped. As management information corresponding to the situation, storing management information indicating a disk array and size of the data, the update list, or the update table in a management information area secured in the disk drive; When the disk array system is restarted, corresponding data is stored in the first disk array system based on the management information stored in the management information area of the disk drive of the first disk array system. Read from the disk drive to the cache memory, 7. The disk array data transfer method according to characterized by comprising the mirror transfer process further comprising the steps of resuming that target data read into Yasshumemori. 11. The method of transferring data between disk arrays according to claim 10, wherein the fixed size is a size of a recording unit of the disk drive. 12. In the update table creating step,
A first update table indicating whether or not there is an update for each large block, for managing the area of the disk drive of the first disk array system by dividing it into large blocks of a first size; Each large block is the first
A plurality of second update tables indicating the presence / absence of updates for each small block for management by dividing into small blocks of a second size smaller than the size of An update block is searched according to an update table, and in the storing step, if the first and second update tables have been created, the first and second update tables are set to the first update table. The management information area is prioritized to be saved in the management information area, and in the restart step, if the first and second update tables are saved in the management information area, the first update table is stored based on the second update table. If only the update table of the first disk array system is stored, the corresponding data is stored in the disk drive of the first disk array system based on the first update table. The disk array data transfer method according to claim 9, wherein a read cache memory. 13. When the first disk array system is stopped, management information regarding data that has been written to the disk drive of the first disk array system and the mirror transfer processing has not been completed is written to the disk drive. A step of storing the management information in the reserved management information area, and a step of saving the management information stored in the management information area of the disk drive of the first disk array system when the first disk array system is restarted. Based on the data, the corresponding data is read from the disk drive of the first disk array system into the cache memory, and the mirror transfer process for the data read into the cache memory is restarted. 2. The disc store according to claim 1, wherein Lee between the data transfer method. 14. A step of detecting that the size of the management information exceeds a predetermined fixed size, and when the size of the management information exceeds the predetermined size,
14. The data transfer method between disk arrays according to claim 13, further comprising the step of switching to the use of new management information in which the corresponding data amount increases but the management information size becomes smaller than the fixed size. 15. The disk array according to claim 14, wherein the fixed size is a size of a memory area reserved for storing the management information or a size of a recording unit of the disk drive. Data transfer method. 16. A mirror disk array system comprising a mirror source disk array system for storing data transferred from a host computer and a mirror destination disk array system for storing a copy of data stored in the mirror source disk array system. In the disk array system applied as the mirror source disk array system, a cache memory for temporarily holding data transferred from the host computer, a group of disk drives for storing data transferred from the host computer, In response to a write request from the host computer, the host side process that receives the data transferred from the host computer and writes the data in the cache memory is executed, and the host side process is completed when the host side process is completed. Host side processing means for returning a completion status for the write request to the storage computer, disk write means for executing disk write processing for writing the data written in the cache memory to the disk drive, and data written in the cache memory To the mirror destination disk array system, and a mirror transfer processing means for performing a mirror transfer processing for performing mirroring independently of the host side processing. 17. A means for detecting the progress status of the mirror transfer processing by the mirror transfer processing means, and the disk write processing by the disk write processing means being stopped according to the detection result of the detection means, and the disk write operation. Release the area of the area on the cache memory where the data that has been written to the disk drive and that has not been subjected to the mirror transfer processing by the processing is written, except for a part of the area where the older data is written. And a releasing means for releasing the data in the partial area on the cache memory in which the data for which the mirror transfer processing has not been completed is emptied by the mirror transfer processing. Of the disk drive of the first disk array system written in the area 17. The disk array system according to claim 16, further comprising a reading unit that reads the written data from the disk drive into the partial area of the cache memory in order to use the data in the mirror transfer processing. . 18. A means for, when the own disk array system is stopped, storing management information on data which has been written in the disk drive and for which the mirror transfer processing has not been completed in a management information area secured in the disk drive. When the own disk array system is restarted, the corresponding data is read from the disk drive into the cache memory based on the management information stored in the management information area, and the data read into the cache memory is read. 18. The disk array system according to claim 17, further comprising a reading unit that restarts the mirror transfer processing by the target mirror transfer processing unit.