JPH09265435A - Storage device system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve storage efficiency without impairing reliability and to improve the relaibility of maintenance processing. SOLUTION: Write data are stored by the format of a RAID (arrangement method for dividing records and storing the divided records in plural disk devices) 3 and cache management information is stored by the format of a RAID (arrangement method (double writing) for storing data having quite the same contents in two disk devices) 1. In the case of a disk controller 1305 constituted of plural clusters 1311, the storage format of data in each cluster is provided with redundancy, and in the case of executing maintenance processing, the storage of new data in a cache memory 1308 for executing maintenance is inhibited and data stored in the cache memory 1308 for executing maintenance are written in a cache memory 1308 in the other cluster 1311 or written in a disk device 1304.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a storage system having a cache memory, and more particularly to high performance and high reliability thereof.

[0002]

2. Description of the Related Art The following Patterson paper is known as a known example of a storage system.

[0003] A. C. M. Sigmod Conference Proceedings, 1988, June, page 109.
−116 (D. Patterson, et al: A Case for Redundant A
rraysof Inexpensive Disks (RAID), ACM SIGMOD confere
nce proceeding, Chicago, IL, June 1-3, 1988, pp.109-11
6) Patterson's paper discloses a technique related to data arrangement on a disk array.

[0004] The disk array is a mechanism for realizing high performance and high reliability of the disk system. In a disk array, a plurality of disk devices are physically made to appear as one logical disk device to a processing device for high performance. On the other hand, for higher reliability, when a failure occurs in a disk device storing data, redundant data for recovering data is stored in another disk device.

The Patterson paper discloses several techniques for arranging redundant data on a disk array. Hereinafter, a method of arranging redundant data related to the present invention will be described.

[0006] The layout method for storing data of exactly the same contents in two disk devices is RAID 1 or
We call it double writing.

A set of data in read / write units when the processing device performs read / write processing with a logical disk device is called a record. Patte
In Rson's paper, the data allocation method of dividing this record and storing it in a plurality of disk devices is called RAID3. In RAID3, 1 for n division units
Redundant data is created.

On the other hand, the data allocation method for storing records in one disk device without dividing the records is RAID4, RA.
It's called ID5. In RAID4 and RAID5, redundant data is created from a plurality of records.

In RAID 1, the same data is stored twice, so the storage capacity of the disk device is required to be doubled.
Therefore, the reliability is improved but the price is increased.

On the other hand, RAID3, RAID4, RAID
In No. 5, the storage capacity does not increase so much because one redundant data needs to be created for a plurality of data.

In RAID 3, a record is divided and recorded in a plurality of disk devices, so that a plurality of disk devices are occupied when the read / write processing requested by the processing device is executed. Before reading / writing data from the disk device, it is necessary to execute seek / search processing. The data read / write time becomes shorter in proportion to the number of divided records, but the seek / search time does not become shorter. Therefore, RAI
In D3, when reading and writing one record, R
Compared to the case where each disk device constituting AID3 is operated independently, the performance is severely degraded. However, in the case of RAID 3, if the amount of read / write data for one time is such that the seek / search time can be ignored, sufficient performance can be obtained. Furthermore, if dedicated hardware is provided, the RAID 3 can be controlled as a single disk device from the disk array control program, so that the disk array control program can be greatly simplified.

In the case of RAID5, a record is stored in one disk device, and therefore when reading one record from the disk device, only one disk device is occupied. However, a new redundant data value is created from the rewritten contents, the value before the update of the record, and the previous value of the redundant data. Therefore, it is necessary to read the value before update of the record and the previous value of the redundant data, write the record and new redundant data, and execute the read / write processing to the disk device four times in total. Also, RAI
In D4 and RAID5, the control software of the disk array is complicated because the disk array control software needs to be aware of each disk device and the processing device needs to make the disk array appear as one logical disk device.

On the other hand, Japanese Unexamined Patent Publication No. 7-230362 discloses a technique relating to a board disc in which a plurality of disc devices are mounted on one board. Normally, the maintenance process is executed on a board basis. A disk array is configured between a plurality of boards so that a read / write request from the processing device can be accepted even during this maintenance processing.

In a recent disk system, a cache memory is provided in a control device for controlling a plurality of disk devices, as disclosed in Japanese Patent Laid-Open No. 5-189314. Furthermore, when the write request is received from the processing device, the control device completes the write request only by writing the write data in the cache memory.
The write data in the cache is written to the disk device by the control device later, but if the cache memory fails before that, the data written to the disk device is lost from the viewpoint of the processing device. For,
It will be a serious situation. For this reason, techniques have been disclosed in which the cache memory is made non-volatile, or the cache data is duplicated to store the write data doubly. In such a case, the management information in the cache memory is also made non-volatile or duplicated.

Further, there is a strong demand for non-stop for the recent disk systems. Especially when performing maintenance processing, while accepting read / write requests from the processing device,
It is necessary to enable maintenance processing. In response to such a request, as seen in Japanese Patent Laid-Open No. 7-306844, the disk system is configured with two clusters, one cluster is blocked and maintenance processing is executed, and the other cluster is used. Some are configured to accept a read / write request from a processing device.

In addition, as a method for improving the performance of the disk system, a technique using a disk cache for performing write-after control is disclosed as follows.

Japanese Patent Application Laid-Open No. 55-157053 discloses a technique relating to speeding up a write request using a write after process in a control device having a disk cache. Specifically, the control device completes the write process at the stage when the write data received from the processing device is written in the cache. Received from the processing device,
The writing of the data stored in the cache to the disk device is later executed by the write after process of the control device.

Japanese Unexamined Patent Publication No. 59-135563 discloses a technique relating to a control device for speeding up write processing while ensuring high reliability.

In Japanese Patent Laid-Open No. 59-135563, a non-volatile memory is provided in the control device in addition to the cache memory.
The write data received from the processing device is stored in the cache memory and the non-volatile memory. The writing of the write data to the disk device is executed by the control device by the write after process. As a result, the reliability of the write after process is improved.

On the other hand, in JP-A-60-114947,
A technique relating to a control device having a disk cache for controlling a double write disk device is disclosed.

In Japanese Patent Laid-Open No. 60-114947, the control device writes the write data received from the processing device into one of the disk device and the cache memory in response to the write request received from the processing device. The control device writes the write data stored in the cache memory later to the other disk device asynchronously with the read / write request from the processing device. The operation in which the control device writes the write data stored in the cache memory to the disk device later asynchronously with the read / write request from the processing device is called a write-after process.

Japanese Unexamined Patent Publication No. 2-37418 discloses a technique for improving the performance of a dual writing disk device by using a disk cache.

In Japanese Patent Laid-Open No. 3-37746, management data of write-after data in a disk cache for the purpose of efficiently executing write-after processing in a controller having a disk cache and executing write-after processing. Techniques for construction are disclosed.

Regarding the disk array, the following techniques have been disclosed.

Japanese Patent Application Laid-Open No. 4-302020 discloses a technique relating to a disk array using a disk device which allows storage of variable length records.

Japanese Unexamined Patent Publication No. 5-46324 discloses a technique for distributing the function of creating an updated value of a parity record among individual disk devices constituting a disk array.

Japanese Unexamined Patent Publication No. 7-44326 discloses a technique for efficiently creating an updated value of a parity record by using two actuators in a disk device or a head capable of executing a read process and a write process in parallel. ing.

[0028]

[Patent Document 1] Japanese Unexamined Patent Publication No. 5-18931
In No. 4, write data is doubly stored in the cache memory. Therefore, the storage capacity of the cache memory needs to be doubled. Similarly, the management information of the cache memory is 2
Since they are duplicated, the required storage capacity is doubled.

That is, in the above-mentioned conventional technique, since the write data and the management information of the cache are simply duplicated, there is a problem that the storage efficiency is poor.

Further, in JP-A-5-189314 and JP-A-7-306844, no consideration is given to making write data redundant during maintenance processing, and if a failure occurs during maintenance processing, write data Etc. may disappear.

A first object of the present invention is to provide a storage system capable of improving storage efficiency without impairing the reliability of write data and cache management information.

A second object of the present invention is to provide a storage system capable of improving the reliability during maintenance processing.

[0033]

According to the present invention, in order to achieve the first object, the write data is RAID
Stored in the format of 3 and the management information of the cache is RAID1.
It is characterized in that it is stored in the format of.

In order to achieve the second object, it is premised that the disk control device is composed of a plurality of clusters. In this case, the storage format of the data in each cluster should have redundancy so that the data to be stored in the cache memory is not duplicated between the clusters. It is characterized in that new data is not stored, and the data that has been stored in the cache memory for maintenance is written to the cache memory of the other cluster or to the disk device.

In the present invention, as described above, in order to improve the storage efficiency without deteriorating the reliability of the write data and the management information of the cache, the write data is RAI.
Store in the D3 format. In this case, the cache memory is usually a maintenance unit for each board.
Configure RAID3. However, this is the only application of RAID3 to the data storage format of the cache memory. Further, the technique of configuring RAID3 between boards is also the application of the board disk configuration method disclosed in Japanese Patent Laid-Open No. 7-230362 to a cache memory.

Therefore, in the present invention, the write data is RA
Stored in ID3 format and cache management information RAI
Store in the D1 format. This is because the management information of the cache has a small unit of update. Therefore, if the management information of the cache is recorded in the RAID3 format, the set of management information that creates redundant data is often partially updated. This is because reading from the cache memory occurs as the information is updated, and the update performance deteriorates. By recording in the RAID1 format, this performance deterioration can be prevented. Further, since the management information of the cache memory is sufficiently smaller than the size of the cache memory, the storage efficiency does not deteriorate so much even if only this part is in the RAID1 format.

On the other hand, since the processing unit updates the write data in units of records, the update unit is large and the write data can be divided into a plurality of cache memories to create redundant data. Moreover, in the case of the cache memory, unlike the disk device, there is no seek search time, so there is no problem in performance even if the write data is divided into a plurality of cache memories. As described above, the write data is changed to RAI.
Store in the D3 format and set the cache management information to RAID
By storing in the format of 1, the storage efficiency can be improved as compared with the related art without significantly impairing the reliability.

Further, in the present invention, as described above, in order to improve the reliability during the maintenance processing, the storage format of the data in each cluster is made redundant, and the data is stored in the cache memory between the clusters. Make sure that the data you use is not duplicated. When performing maintenance processing, the cache memory for maintenance does not store new data, and the data previously stored in the cache memory for maintenance is written to the cache memory of the other cluster. Or write to the disk device. The data in the cache memory that is not maintained also has redundancy. This makes it possible to improve reliability during maintenance processing.

[0039]

Embodiments of the present invention will be described below with reference to the drawings.

First, the first embodiment will be described.

FIG. 1 is a block diagram showing an embodiment of a disk array control processor to which the present invention is applied, and is a block diagram of a computer system using the disk array control processor of FIG.

The computer system shown in FIG.
300, a control device 1305, and one or more disk devices 1304.

The processing unit 1300 comprises a CPU 1301, a main memory 1302 and a plurality of channels 1303.

The controller 1305 follows the read / write request from the processor 1300.
0 and the disk device 1304 execute data transfer processing, and are provided with two clusters 1311.

Each cluster 1311 includes one or more disk array control processors 1310, a cache memory (hereinafter, simply referred to as cache) 1308, a bus 1312, a power supply 1312, and an alternate power supply 1313.
Bus 1312 is connected to the disk array control processor 1310 and cache 1308 of the opposite cluster 1311.

The power supply 1313 supplies power to the cluster 1311, while the alternate power supply 1314 supplies power when a failure occurs in the power supply 1313. That is, the alternate power supply 1314 is a standby system for the power supply 1313. In FIG. 2, the alternate power supply 1314 is a standby system for the power supply 1313 of the same cluster 1311, but it may be a standby system for the opposite cluster 1311.

In the cache 1308, the disk device 1
A part of data in 304, management information of this data, management information of the control device 1305, etc. are stored.
From the viewpoint of reliability, it is desirable to make nonvolatile.

Disk Array Control Processor 1310
Has a function of receiving a read / write request from the processing device 1300 in the control device 1305 and utilizing the cache 1308 or the like to execute a data transfer process between the processing device 1300 and the disk device 1304.

This disk array control processor 131
0 is composed of a read / write accepting unit 100, a write-after process executing unit 101, a block migration process executing unit 102, and a recovery process executing unit 103. The operation of each of these units will be described later with reference to the flowchart.

Hereinafter, in the present embodiment, the control device 13
The operation when 05 receives a read / write request from the processing device 1300 will be briefly described. The operation described below is executed by the disk array control processor 1310 in the control device 1305.

First, the control unit 1305 is the processing unit 130.
The operation when a read request is received from 0 will be described.

When receiving a read request from the processing device 1300, the control device 1305 transfers the record 500 in the cache 1308 to the processing device 1300 if the target record 500 (FIG. 5) is stored in the cache 1308. And complete the operation.

However, the record 500 to be read
Is not stored in the cache 1308, the record 500 is read from the disk device 1304 and transferred to the cache 1308. Thereafter, this record 500 is transferred from the cache 1308 to the processing device 1300, and the operation is completed.

On the other hand, the control unit 1305 is the processing unit 130.
When a write request from 0 is received, the disk device 1
The operation is completed when the write data to be written in the record 500 on the 304 is stored in the cache 1308. The write of the write data to the disk device 1304 is executed by the control device 1305 later. This process is called a write-after process. When the write-after process is executed, if the disk array configuration is RAID 3 or RAID 5, parity data corresponding to the record 500 is created and written to the disk device 1304. In the case of the RAID 1 configuration, the contents of the record 500 are written in the two disk devices 1304.

FIG. 3 shows the cache 1 in this embodiment.
It is a figure which shows the logical structure of 308. Cache 13
08 is composed of a management information storage area 300 and a disk data storage area 301. Management information storage area 300
Is the disk device 130 stored in the cache 1308.
4 stores data management information and management information of the control device 1305. The disk data storage area 301 stores a part of data in the disk device 1304.

FIG. 4 is a diagram showing a physical structure of the cache 1308. The cache 1308 includes a plurality of boards 4.
00. The plurality of boards 400 are roughly classified into a management information board 401 and a disk cache board 40.
And 2.

The management information board 401 is the board 400 corresponding to the management information storage area 300, and the disk cache board 402 is the board 400 corresponding to the disk data storage area 301. The unit of failure detection is the unit of the board 400.

FIG. 5 shows the disk cache board 402.
FIG. 5 is a diagram showing a storage format of data in a record 500.
Is a read / write unit of the disk device 1305.
The record 500 is divided into m pieces (m ≧ 2), and each piece is stored as a division record 501 in the disk cache board 402 corresponding to each division unit.

In addition, n pieces (n ≧
The redundant data 502 of 1) is created. The m divided records 501 and the n redundant data 502 are separately stored in m + n disk cache boards 402, respectively.

However, in the first embodiment, the record 50
It is sufficient if 0 has redundancy in the storage format. For example, the contents of the record 500 are duplicated in the disk cache board 402.
May be stored in.

Incidentally, one disk cache board 4
02 may store only the divided record 501 or only the redundant data 502. Alternatively, the divided record 501 and the redundant data 502 may be mixed and stored in one disk cache board 402. When n redundant data 502 are stored, the contents of the record 500 can be restored even if n disk cache boards 402 fail.

FIG. 6 is a diagram showing a storage format of information of the management information board 400. The management information 600 is doubly stored in the two management information boards 401.

The two management information boards 401 may be paired and the same management information 600 may be stored in the two management information boards 401. Also, management information 600
May be stored in two arbitrary management information boards 401, and the information of the entire management information board 401 may be different. Further, in the first embodiment, the storage format of the management information 600 does not necessarily have to be stored in duplicate, and may be redundant.

In this embodiment, when a failure occurs in the board 400 and the redundancy of the storage of the management information 600 and the record 500 is lost, the preprocessing for closing the cluster 1311 is started.

That is, the use of the cache 1308, which has lost redundancy, is to be stopped. Immediately, the reason why the block cannot be blocked is that the cache 1308 to be blocked stores write data that has not been written to the disk device 1304.

The preprocessing for closing the cluster 1311 is
The write data stored in the cache 1308 to be blocked and not written in the disk device 1304 is written in the disk device 1304, or the opposite cluster 13
11 is a process for copying to the cache 1308. Similarly, in this embodiment, when the cache 1308 is to be maintained, the entire cache 1308 of one cluster 1311 is to be maintained.

As described above, in the present embodiment, when the caches 1308 lose their redundancy, the caches 1308 with redundancy are immediately used in a concentrated manner. Can be secured.

Similarly, when the cache 1308 of one cluster 1311 is maintained, the remaining clusters 13
Since the 11 caches 1308 have redundancy, the reliability can be secured even if the write after process is performed.

FIG. 7 summarizes the state transitions of the cache 1308. In the present embodiment,
Since there are two caches 1308, this state is managed for each cache 1308.

In FIG. 7, the available state 900 indicates that the cache 1308 is available. When a failure of the board 400 of the relevant cache 1308 is detected from the available state 900 or when the maintenance process is started, the relevant cache 1308 is in the blocking transitioning state 901.
To This means that the preprocessing for closing the cache 1308 is being executed. When this processing is completed, the state of the cache 1308 changes to the closed state 902.
become.

After that, when the maintenance of the cache 1308 is completed, or the replacement of the failed board 400 is completed, the state of the cache 1308 is restored to the recovery state 90.
It becomes 3. This represents a state in which pre-processing for bringing the cache 1308 into the usable state 900 is being executed.

FIG. 8 is a diagram showing the logical structure of the disk data storage area 301 shown in FIG. 3, and the disk data storage area 301 is composed of a plurality of segments 800. A record 500 is stored in the segment 800.

FIG. 9 shows a record 500 and a segment 80.
The relationship between 0 and the disk array board 400 is shown. The segment 800 is divided into m + n divided segments 910, and a divided record 501 and redundant data 502 are stored in each divided segment 901.

FIG. 10 is a diagram showing the logical structure of the management information storage area 300 shown in FIG. 3. In this embodiment, the information contained in the management information storage area 300 is the cache state 1000 in the local cluster. A cache state 1001 in another cluster, segment management information 1002 corresponding to a segment, and a free segment queue pointer 1003.

The self-cluster cache status 1000 and the other-cluster cache status 1001 are the cluster 1 of the own cluster.
The cache 1308 of 311 or the cache 1308 of another cluster 1311 is in each of the states shown in FIG. 7.

The segment management information 1002 is management information existing corresponding to each of the segments 800 in FIG. The free segment queue pointer 1002 is, as shown in FIG. 11, the segment management information 1002 corresponding to the segment 800 that does not store the record 500.
This is a pointer indicating the beginning of the empty segment queue 1100 in which are combined.

These local cluster cache states 10
00, cache status in other cluster 1001, segment management information 1002, free segment queue pointer 1
Any information 003 is duplicated and stored in the two management information boards 401.

FIG. 12 is a diagram showing the details of the segment management information 1002. The segment management information 1002 is
Segment pointer 1200, in-use bit 1201,
The structure includes a dirty bit 1202 and a disk address 1203.

The segment pointer 1200 stores the segment management information 100 in the free segment queue 1100.
This is for pointing to the next segment management information 1002 when 2 is connected to the empty segment queue 1100.

The in-use bit 1201 is turned on while each execution unit in the disk array control processor 1310 is executing the read / write processing of the record 500 stored in the segment 800 corresponding to the segment management information 1002. That is, when the read / write process acceptance unit 100, the write-after process execution unit 101, the block migration process execution unit 102, the recovery process execution unit 103, etc. use the segment 800, this bit 1201 is turned on. When the other execution unit finds the segment management information 1201 in which the in-use bit 1201 is on, it waits until it is turned off. This makes it possible to exclude each process.

As for the dirty bit 1202, the contents of the record 500 stored in the segment 800 corresponding to the segment management information 1002 are still the disk device 130.
4 indicates that the record 500 is not written, that is, the record 500 needs to be subjected to the write-after process.

The disk address 1203 corresponds to which disk device 1304 the record 500 stored in the segment 800 corresponding to the segment management information 1002 indicates.
And the address in the disk device 1304.

The processing executed by the disk array control processor 1310 in the control device 1305 according to this embodiment will be described below with reference to the flow charts of FIGS.

When any of the caches 1308 of the two clusters 1311 is in the usable state 900 or the recovering state 903, which of the caches 1308 is to be used depends on the disk of the record 500 to be read / written. It is assumed that it is uniquely determined by the address of the device 1304.

FIG. 13 is a flowchart showing the processing of the read / write accepting unit 100. This processing is started when a read / write request is accepted from the processing device 1300.

First, at step 110, when either cache 1308 is in the usable state 900, it is determined which cache 1308 is to be used. Next, in step 111, the state of the cache 1308 is checked.
In the case of the usable state 900, the cache 1308 is used as it is, and the read / write processing is started in step 112.

However, in the closed state 902, it is decided in step 113 to use the opposite cache 1308, and then the process jumps to step 112.

Further, in the case of the blocking transfer in progress state 901, first, in step 114, it is checked whether or not the record 500 to be read / written is stored in the cache 1308. If not stored, the process jumps to step 113. If so, in step 115, the contents of this record 500 are reversed to cache 1
It is copied to 308, or it is checked by the dirty bit 1202 in FIG. 12 whether the contents of this record 500 have not been written in the disk device 1304. If not, it is written in the disk device 1304.

When writing to the disk device 1304, if the disk array configuration is RAID 3 or RAID 5, parity data corresponding to the record 500 is created and written to the disk device 1304. In the case of the RAID 1 configuration, the contents of the record 500 are written in the two disk devices 1304.

Further, in step 116, this record is deleted from the cache 1308. After this, the process jumps to step 113.

In the case of the recovering state 901, step 117
Then, it is checked whether or not the record 500 to be read / written is stored in the cache 1308 on the opposite side. If not stored, the process jumps to step 112. If so, in step 118, this record 5
00 is copied from the opposite cache 1308 to the cache 1308, or it is checked whether the content of this record 500 has been written to the disk device 1304. If not, the disk device 130
Write to 4.

When writing to the disk device 1304, the disk array configuration is RAID3 or RAI.
In the case of D5, parity data corresponding to the record 500 is created and written in the disk device 1304. RAID
In the case of the configuration of 1, the content of the record 500 is written in the two disk devices 1304.

Further, in step 119, this record is deleted from the cache 1308. After this, read / write processing starts.

FIG. 14 shows the write after process execution unit 101.
Is a flow chart showing the processing of the above, the outline of which is to write the unwritten data in the cache 1308 to the disk device 1304, as already described (step 131).

FIG. 15 is a flowchart showing the processing of the blocking migration processing execution unit 102. This processing is started when a failure of the board 400 of the cache 1308 is detected or when the maintenance processing is started. This processing is
This is a pre-process for closing the cache 1308.

Specifically, in step 120, it is checked whether or not there is a record 500 stored in the cache 1308. If found, in step 121, it is checked whether the content of this record 500 is copied to the opposite cache 1308 or whether this record 500 needs to perform write-after processing, and write-after processing needs to be performed. If it is data, the content is written in the disk device 1304.

When writing to the disk device 1304, if the disk array configuration is RAID 3 or RAID 5, parity data corresponding to the record 500 is created and written to the disk device 1304. In the case of the RAID 1 configuration, the contents of the record 500 are written in the two disk devices 1304.

Thereafter, this record 500 is deleted from the cache 1308. When there are no more records 500 stored in the cache 1308, step 1
22, the state of the cache 1308 is closed 90
Then, the process is completed.

FIG. 16 is a flow chart showing the processing of the recovery processing execution unit 103. This processing is started when the maintenance of the cache 1308 concerned is completed, or the replacement of the failed board 400 is completed. It This process is a pre-process for bringing the cache 1308 into the usable state 900.

First, in step 123, it is checked whether or not there is a record 500 to be stored in the cache 1308 in the opposite cache 1308. If found, in step 124, the content of this record 500 is copied to the cache 1308, or if this record 500 is the record 500 to be write-after,
The contents are written in the disk device 1304.

When writing to the disk device 1304, if the configuration of the disk array is RAID3 or RAID5, parity data corresponding to the record 500 is created and written to the disk device 1304. In the case of the RAID 1 configuration, the contents of the record 500 are written in the two disk devices 1304.

After that, the record 500 is erased from the opposite cache 1308. Opposite cache 130
8 records 50 to be stored in the cache 1308
When 0 disappears, in step 125, the state of the cache 1308 is set to the usable state 902, and the processing is completed.

Here, the read / write processing acceptance unit 100
The processes of the write-after process execution unit 101 and the block transfer process execution unit 102 are executed in parallel. The exclusion of each process is executed by using the usage bit 1201 in the segment management information 1002 of the cache 1308. Similarly, the read / write processing acceptance unit 100 and the write-after processing execution unit 1
01 and the process of the recovery process execution unit 103 are also executed in parallel. Also in this case, exclusion of each process is performed by the cache 1308.
Is executed using the use bit 1201 of

In the present embodiment, the storage format of the data in each cluster has redundancy such as RAID3 so that the data stored in the cache memory is not duplicated between the clusters. Furthermore, when performing maintenance processing, do not store new data in the cache memory that performs maintenance, and use the data that was previously stored in the cache memory that performs maintenance in the cache memory of the other cluster. I tried to write or write to the disk device. As described above, there is an effect that reliability during maintenance processing can be improved.

Next, a second embodiment of the present invention will be described.

FIG. 17 is a configuration diagram showing a second embodiment of a computer system to which the present invention is applied.

The structural difference between the second embodiment and the above-described first embodiment is as follows.

(1) The control unit 1305 is the cluster 131
Do not have 1.

(2) Controller 1305 of the second embodiment
Has two buses 1312.

(3) The configuration of the management information storage area 300 of the cache 1308 is different.

Management information storage area 3 in this embodiment
The configuration of 00 is shown in FIG. The management information storage area 300 of this embodiment includes a cache status 1400, segment management information 1002 corresponding to a segment, and an empty segment queue pointer 1003. The difference from the first embodiment is that since there is only one cache 1308, the cache 1308 has only one state.

Other than this, FIG. 3 to FIG. 6, FIG. 8 to FIG.
The configurations of 1 to 12 can also be applied to the second embodiment.

However, in this embodiment, the management information 600 is always stored in the two management information boards 401 as shown in FIG. Further, in this embodiment, as shown in FIG.
The record 500 is always stored in each disk cache and the board 402 in the RAID3 format.

In this embodiment, the cache 1308 is 1
Since there is only one, the unit of blockage of the cache 1308 at the time of failure is the board 400. Similarly, the cache 130
The maintenance unit of 8 is also the board 400.

FIG. 19 summarizes the state transition of the cache 1308 in this embodiment. Normal state 1
500 is a state in which the cache 1308 has redundancy. In this state, the control device 1305 controls the processing device 1300.
When the write request is received from, the process is completed when the write data is written in the cache 1308.

When a failure occurs in the board 400 of the cache 1308 or when maintenance processing is executed, the cache 1308
When there is no redundancy in 308, there is no redundancy transition state 15
Move to 01. In the non-redundancy transition state 1501, the contents of the record 500 in the cache 1308 are not written to the disk device 1304, but the contents are written to the disk device 130.
4 is written. When this is completed, the state of the cache 1308 transits to the no redundancy state 1502.

When the cache 1308 is in a state other than the normal state 1600, from the viewpoint of reliability, the control device 1305 completes the process when the write data is written to the disk device 1304 when the write request is received. .

Incidentally, the control device 130 for the read request.
5 does not depend on the state of the cache 1308.
This is the same as in the embodiment.

FIG. 20 is a diagram showing the configuration of the disk array control processor 1310 in the control device 1305, which is composed of a read / write accepting unit 1600, a non-redundant migration process executing unit 1601, and a write-after process executing unit 101. ing.

The processing executed by the disk array control processor 1310 will be described below with reference to the flowcharts of FIGS. 21 and 22. The write-after process execution unit 101 is the same as that in the first embodiment, and therefore its description is omitted.

FIG. 21 shows a read / write acceptance unit 1600.
2 is a flowchart showing the processing of FIG. 1, and the read / write accepting unit 1600 starts its operation when accepting a read / write request from the processing device 1300.

First, in step 1610, it is determined whether the read request or the write request is received. If it is a read request, in step 1611, the read process is started. If it is a write request, in step 1612,
The state of the cache 1308 is analyzed.

If the cache is in a normal state, step 1
At 613, the write operation of writing the write data received from the processing device 1300 to the cache 1308 is completed.
In the case of the no redundancy state 1502, in step 1614, the write data received from the processing device 1300 is written to the disk device 1304, and the operation is ended.

In the case of the transition state 1501 without redundancy, FIG.
The processing of the non-redundancy migration processing execution unit 1601 shown in 2 is started.

In this redundancy-free migration processing, step 16
15, the content of the record 500 is the disk device 1304.
If it is written, the process jumps to step 1617 to update the cache state to the no redundancy state 1502. If not, the unwritten record 5
00 is written in the disk device 1305. Then, the process jumps to step 1615.

Note that step 1614 and step 1
In 616, when writing the unwritten record 500 to the disk device 1304, the configuration of the disk array is RAID.
In the case of 3 or RAID 5, parity data corresponding to the record 500 is created and written in the disk device 1304. In the case of the RAID 1 configuration, the contents of the record 500 are written in the two disk devices 1304.

After that, when there is no record 500 in the cache 1308 whose contents are not written in the disk device 1304, the state of the cache 1308 is updated to the non-redundancy state 1502 in step 1617, and the processing is completed.

As is clear from the above description, according to the present embodiment, the write data is stored in the RAID3 format and the cache management information is stored in the RAID1 format, so that the reliability is greatly impaired. Without
The storage efficiency can be improved as compared with the conventional one.

[0129]

As described above, in the present invention, the storage format of the write data of the disk cache is set to the RAID3 format and the cache management information is stored in the RAID1 format, so that the reliability is not impaired. The storage efficiency can be improved as compared with the conventional one.

On the other hand, according to the present invention, in the case of a disk control device composed of a plurality of clusters, the storage format of the data in each cluster is made redundant, and further maintenance processing is performed. Do not store new data in the cache memory that performs the operation, and write the data that was stored in the cache memory that performs maintenance to the cache memory of the other cluster, or
I tried to write to the disk device. As a result, it is possible to improve the reliability during the maintenance process.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a first embodiment of a disk array control processor to which the present invention is applied.

FIG. 2 is a configuration diagram of a computer system using the disk array control processor of FIG.

FIG. 3 is a logical configuration diagram of a cache memory.

FIG. 4 is a physical configuration diagram of a cache memory.

FIG. 5 is a diagram showing a storage format of a record of a disk cache board.

FIG. 6 is a diagram showing an information storage format of a management information board.

FIG. 7 is a diagram showing a state transition of a cache memory in the first embodiment.

FIG. 8 is a logical configuration diagram of a disk data storage area.

FIG. 9 is a diagram showing a physical storage format of a segment.

FIG. 10 is a logical configuration diagram of a management information storage area in the first embodiment.

FIG. 11 is an explanatory diagram of an empty segment queue.

FIG. 12 is a configuration diagram of segment management information.

FIG. 13 is a flowchart showing a process of a read / write accepting unit in the first embodiment.

FIG. 14 is a flowchart showing a process of a write after process execution unit in the first embodiment.

FIG. 15 is a flowchart showing processing of a blocking migration processing execution unit according to the first embodiment.

FIG. 16 is a flowchart showing processing of a recovery processing execution unit in the first embodiment.

FIG. 17 is a configuration diagram showing a second embodiment of the present invention.

FIG. 18 is a logical configuration diagram of a management information storage area in the second embodiment.

FIG. 19 is a cache 1308 according to the second embodiment.
It is a figure which shows the state transition of.

FIG. 20 is a configuration diagram of a disk array control processor according to the second embodiment.

FIG. 21 is a flowchart showing a process of a read / write accepting unit in the second embodiment.

FIG. 22 is a flowchart showing processing of a non-redundancy migration processing execution unit according to the second embodiment.

[Explanation of symbols]

Reference numeral 100, 1600 ... Read / write acceptance unit, 101 ... Write after processing execution unit, 102 ... Block migration processing execution unit,
103 ... Recovery processing execution unit, 300 ... Management information storage area,
301 ... Disk data storage area, 400 ... Board, 4
01 ... Management information board, 402 ... Disk cache board, 500 ... Record, 501 ... Divided record, 50
2 ... Redundant data, 800 ... Segment, 901 ... Divided segment, 1300 ... Processing device, 1304 ... Disk device, 1305 Control device, 1308 ... Cache memory,
1310 ... Disk array control processor, 1311 ...
Cluster, 1312 ... Bus, 1313 ... Power supply, 1314
... Alternate power supply, 1601 ... Non-redundant migration processing execution unit.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical display location G06F 12/08 310 7623-5B G06F 12/08 310Z 320 7623-5B 320 (72) Inventor Yoshihiro Azumi 2880 Kozu, Odawara-shi, Kanagawa Stock Company Hitachi Storage Systems Division

Claims (16)

[Claims] 1. A storage device system comprising one or more storage devices and a control device having a plurality of clusters, wherein each cluster has a redundant cache memory, and the control device has each cluster Means for storing the data in the storage device in the cache memory of each cluster when the cache memory is in a normal state; and a write request accepted from the processing device in response to the write request from the processing device. A storage device system, which stores data in the cache memory and completes the write request. 2. As a pre-process when the maintenance of the cluster is performed, the write data not yet written in the storage device in the cache memory in the maintenance target cluster is stored in the cache memory of the cluster not performing maintenance or the write data. 2. The storage device system according to claim 1, further comprising means for writing in the storage device. 3. A unit for writing write data, which has not been written in the storage device in the cache memory in the maintenance target cluster, to the cache memory or the storage device in the cluster that does not perform maintenance, and read from the processing device. A write request is received, and a read / write process is executed in parallel with the write process of the write unit.
The storage system described. 4. After the maintenance of the maintenance target cluster is completed, the write data not yet written in the storage device in the cache memory other than the maintenance completed cluster is replaced with the cache memory in the maintenance completed cluster. 3. The storage device system according to claim 2, further comprising means for writing in the storage device. 5. Means for writing write data that has not yet been written to the storage device in a cache memory other than the maintenance completed cluster to the cache memory or the storage device of the maintenance completed cluster. 5. A means for receiving a read / write request from the processing device, and executing the read / write processing in parallel with the write processing of the write means.
The storage system described. 6. When the cache memory in the cluster loses redundancy due to a failure, the write data not yet written to the storage device in the failed cache memory is replaced by the failed cache. The storage system according to claim 1, further comprising a cache memory other than the memory or a unit for writing in the storage. 7. Write data that has not yet been written to the storage device in the failed cache memory,
A means for writing to a cache memory other than the failed cache memory or the storage device, and means for receiving a read / write request from the processing device and executing the read / write processing in parallel with the write processing of the writing means. 7. The method according to claim 6, wherein
The storage system described. 8. After the redundancy of the cache memory is restored, the write data not yet written in the storage device in the cache memory other than the cache memory where the redundancy is restored is stored in the cache memory where the redundancy is restored or 7. The storage device system according to claim 6, further comprising means for writing in the storage device. 9. A means having means for writing write data, which has not been written in the storage device, in a cache memory other than the cache memory in which redundancy has been restored, to the cache memory in which redundancy has been restored or the storage device. 5. A means for receiving a read / write request from the processing device, and executing the read / write processing in parallel with the write processing of the write means.
The storage system described. 10. A storage device system comprising one or more storage devices and a control device, comprising a plurality of boards, a cache memory for storing a part of data of the storage devices, and a plurality of boards. And a management information storage memory for storing management information of data stored in the cache memory, and a part of data of the content of the storage device is divided into m (≧ 2) boards in the cache memory. Means for storing redundant data for recovering the contents of the data in n (≧ 1) boards in the cache memory, and storing the management information in two or more pieces in the management information memory. Storage system having means for redundantly storing the same on the board. 11. When both the cache memory and the management information storage memory have redundancy, a write request is received from a processing device, and write data received from the processing device in association with the write request is written to the cache memory. 11. The storage device system according to claim 10, further comprising means for storing the data in a storage medium and completing the write request. 12. A write request accepted from a processing device when a redundancy is lost due to a failure of either the cache memory or the management information storage memory, and the write request accepted from the processing device in association with the write request. 11. The storage device system according to claim 10, further comprising means for completing the write request after writing data in the storage device. 13. When any one of the cache memory and the management information storage memory loses redundancy due to a failure, write data not yet written in the storage device in the cache memory is written in the storage device. 13. The storage device system according to claim 12, further comprising means. 14. A means for writing write data, which has not been written in the storage device in the cache memory, to the storage device as a pre-process when the board of the cache memory or the management information storage memory is maintained. The storage device system according to claim 10, further comprising: 15. A write request is received from a processing device when the board of the cache memory or the management information storage memory is being maintained, and the write data received from the processing device in association with the write request is stored in the storage device. 15. The storage device system according to claim 14, further comprising means for completing the write request after writing to the device. 16. A unit having a unit for writing write data, which has not yet been written to the storage device in the cache memory, to the storage device, and a write request from a processing device, the processing being accompanied by the write request. The storage device according to claim 15, further comprising: a unit that, after writing the write data received from the device to the storage unit, executes a process of completing the write request in parallel with the writing process of the writing unit. system.