JPH08123627A - Disk array device - Google Patents

Abstract

PURPOSE: To provide a disk array device which can decreases the redundancy when data has redundant constitution within a range wherein a necessary throughput is obtained. CONSTITUTION: The read time when data are decentralized to and stored on plural disk devices 12-15 and the read time when compressed data are decentralized to and stored on the disk devices 12-14 except one specific device are found respectively. When the read time is shorter than that when the uncompressed data are decentralized to and stored on all the disk devices as a result of compression although the number of the disk devices which are read at the same time in parallel is one less, the data after compression are put in RAID constitution and stored. Consequently, the throughput of the read never becomes lower than that at the time of doubled constitution. The data in the RAID constitution are stored, the redundancy is reducible on condition that the necessary throughput.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array device, and more particularly to a disk array device for storing data in a redundant configuration.

[0002]

2. Description of the Related Art A disk device is often used as an auxiliary storage device in an information processing device such as a host computer. The problems that the disk device has are that the time required for reading and writing data is longer than the processing speed of the information processing device, and the stored data may be damaged. Among these, there is a disk array device as a disk device in which the time required for reading and writing is shortened. This device stores data to be stored in a plurality of disk devices in a distributed manner, and at the time of reading, the data is simultaneously read from these disk devices in parallel. Throughput is improved by reducing the amount of data read from one disk device.

Further, there is a disk device in which the throughput is improved by reducing the amount of data actually read from and written to the disk device. In this device, the data to be stored is compressed to reduce the amount of the data, thereby improving the apparent throughput. A disk device provided with the function of performing such data compression and expansion in the disk device itself is disclosed in Japanese Patent Laid-Open No. 4-364547. Further, Japanese Patent Laid-Open No. 1-237966 discloses a disk device in which data is compressed for each predetermined amount of data and blank space data is written in an empty area generated by the compression. In this device, a storage area on the disk is secured in accordance with the amount of data before compression, and when the amount of data is reduced by compression, predetermined blank space data is written in the empty area created in the secured area. When reading the compressed data, the reading time is shortened by ending the reading of the data when the margin data is detected or by permitting the head to move to another track.

Japanese Unexamined Patent Publication No. 1-58547 discloses a disk device in which a storage area is divided into a first area for storing compressed data and a second area for storing uncompressed data. Data is preferentially written in the second area, and when this area becomes full, the oldest data is compressed and moved to the first area. If the target data is in the second area, it is not necessary to decompress the data in the reading process, so the reading time is shortened accordingly. In addition, the storage capacity of the disk device is apparently increased because the storage area is provided for compression.

Further, Japanese Patent Application Laid-Open No. 2-108119 discloses a disk device in which data is compressed by a plurality of compression systems and the compression system having the highest compression rate is selected. In this disk device, when the amount of data increases as a result of compression as compared with that before compression, the data is stored in the disk device in an uncompressed state.

On the other hand, there is a disk device which stores data in a redundant configuration in order to prevent the data from being damaged and improve the reliability of the disk device. Among them, the mirrored disk device is provided with two disk devices and stores the same data in them. Then, when one of the disk devices fails, the data is read from the other disk device. Such a redundant configuration in which the same data is stored in two disk devices is called a dual configuration. In addition, there is a system in which three or more disk devices are prepared, and one of them stores parity data for error correction to reconstruct the data of the damaged disk device. Data reconstruction technology using parity data is generally known as RAID (Redundant Arrays of Inexpensiv).
It is called the third level of e Disks technology, and is particularly suitable for storing a large amount of data in a redundant configuration.

FIG. 23 shows a data flow in a disk device which reconstructs data damaged by parity data. The first disk device and the second disk device are adapted to store data to be written. The third disk device is adapted to store parity data. The storage areas of these disk devices are divided into blocks each having a size of 4 kilobytes, and the blocks are associated with each other on a one-to-one basis among the disk devices. For example, when storing data for two blocks, the data is distributed and stored in the associated blocks 201 and 203 of the first disk device and the second disk device. Then, the block 201 of the first disk device and the block 2 of the second disk device
Exclusive OR is taken bit by bit between 02 data,
The result is stored in the block 203 of the third disk device. First, the leading 1-byte data 204 and 205 of the block 201 and the block 202 are respectively read out, the exclusive OR is taken for each bit, and the parity data 206 as the operation result is the block of the third disk device. It is stored at the head of 203. This is sequentially performed byte by byte from the beginning of the block. Hereinafter, such a redundant data structure will be referred to as a RAID structure.

Here, it is assumed that the first disk device has failed at the time of reading data. In this case, the data of the block 201 is reconstructed by reading the blocks 203 and 202 of the third disk device and the second disk device and taking the exclusive OR of these data.
For example, the parity data 20 at the beginning of the block 203
6 and the head data 205 of the block 202 are exclusive-ORed bit by bit to reconstruct the head data 204 of the block 201.

[0009]

When data is stored in a duplicated structure as in a mirrored disk device, the amount of data to be stored is doubled and its redundancy is large, so that data that can be stored in the disk device is stored. There is a problem that the amount is significantly reduced. The redundancy in the RAID configuration is less than that in the duplex configuration. For example, when four disk devices are used, data is distributed and stored in three disk devices, and parity data is stored in the remaining one. Therefore, it is possible to store the data in a redundant configuration with a storage area that is four thirds of the amount of data to be written. However, as compared with the case where the same four disk devices are used, the data read time becomes longer than in the case of the dual configuration. This is because the number of disk devices that read data simultaneously in parallel decreases from four to three. Therefore, in order to secure the throughput at the same level as the case of the duplex configuration when storing data in the RAID configuration,
There is a problem that the number of disk devices must be increased by one. On the contrary, when the number of disk devices is limited, there is a problem that the throughput is reduced as compared with the case of the dual configuration.

As a method of improving the read throughput in the case of the RAID configuration, the compressed data is converted into R
An AID configuration can be considered. If the amount of data is reduced by compression, the amount of data read from the disk device per unit is reduced, and the required reading time can be shortened accordingly. However, the compression rate varies depending on the compression algorithm and the data to be compressed, and even if the compression is performed, it may not be possible to ensure the same level of throughput as in the duplex configuration. Therefore, when it is used as an auxiliary storage device of an information processing device such as an image processing device that needs to read data within a predetermined time limit, the number of disk devices must be increased in consideration of the case of poor compression ratio. There is a problem of not becoming. As described above, the redundant configuration has a problem that the redundancy of data is large and the amount of data that can be stored in the disk device is reduced.
The ID configuration has a problem that the throughput is lower than that of the duplex configuration.

Further, when the data is stored in the disk device in the redundant configuration or the RAID configuration, the time required for writing becomes longer than in the case where the data is stored without the redundant configuration. For example, in the case of the RAID configuration, it is necessary to take the exclusive OR of each bit of the data to be stored and store this in the disk device prepared for the parity data. Therefore, there is a problem that it takes a long time to release the information processing apparatus that has issued the data write command from the write processing.

Therefore, a first object of the present invention is to provide a disk array device capable of storing data in a redundant configuration and releasing an information processing device which issues a write command from the writing process in a short time. It is in.

A second object of the present invention is to provide a disk array device which can reduce the redundancy of data within a range where a required throughput can be obtained.

[0014]

According to a first aspect of the present invention, a plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and input from the write data input means. Distributed storage means for distributing and storing the stored data to a plurality of disk devices, and sending a predetermined writing completion notification indicating the end of writing when the storage of the data by the distributed storage means is completed. And a redundant configuration storage unit configured to redundantly store the data stored by the distributed storage unit after the writing completion notification is transmitted by the writing completion notification transmission unit and store the redundant data in a plurality of disk devices. Prepared for.

That is, according to the first aspect of the present invention, a write completion notice is sent at the time when the data to be written is distributed and stored in a plurality of disk devices, and then the data is re-stored in a redundant configuration. There is. As a result, the host device that has issued the write command can be released from the writing process in a short time.

According to a second aspect of the present invention, a plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and data input from the write data input means for a plurality of disk devices. Distributed storage means for distributing and storing to devices, and writing completion notification sending means for sending a predetermined writing completion notification indicating the end of writing when the storage of data by the distributed storage means is completed,
Redundant configuration storage means for storing the data stored by the distributed storage means in a redundant configuration and re-storing the data in a plurality of disk devices after the write completion notification is transmitted by the write completion notification transmission means, and a data read command And a data stored in the distributed storage means if the data to be read at the time when the read command is received by the read command receiving means is redundantly configured by the redundant configuration storage means and is not stored again. The disk array device is equipped with a reading means for reading.

That is, according to the second aspect of the present invention, when the data to be written is distributed and stored in a plurality of disk devices, a write completion notice is sent out, and then the data is re-stored in a redundant configuration. There is. When a read command is received before the re-storing process is completed, the data is initially read from the dispersedly stored areas. As a result, the data can be read even during the process of re-storing the data in the redundant configuration.

In a third aspect of the present invention, a plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and data input from the write data input means are stored in the disk device. First read time indexing means for obtaining the read time required to read this data from these disk devices simultaneously in parallel when distributed and stored, and compression for compressing the data input from the write data input means. Means and the data after being compressed by the compressing means are distributed and stored in a disk device other than a predetermined one of a plurality of disk devices, the data after the compression are simultaneously processed in parallel from these disk devices. A second read time indexing means for determining a read time required for reading and a read time for the second read time determined by the first read time indexing means. When the read time obtained by the time indexing means is shorter than or equal to the read time, the first distributed storage means stores the data input from the write data input means in a distributed manner in the disk device, and the input from the write data input means. Second distributed storage means for storing the distributed data by changing the first distributed storage means and the disk device of the storage destination to store the data, and the read time first read by the second read time indexing means. Third distributed storage means for storing the data after being compressed by the compression means in a disk device other than a predetermined one when the read time is shorter than the read time obtained by the time indexing means, and the third distributed storage. Any one of disk devices in which data is stored by means
When the table becomes unreadable, an error correction code generating means for generating an error correction code for reconstructing the data stored therein and an error correction code generated by this error correction code generating means are set to a predetermined value 1. The disk array device is provided with an error correction code storage means for storing in one disk device.

That is, according to the third aspect of the invention, the read time when data is distributed and stored in a plurality of disk devices, and the compressed data is distributed and stored in disk devices other than a predetermined one. Read time of each is calculated. Even if the number of disk devices that are read in parallel at the same time is reduced by one, when the read time becomes shorter than that when uncompressed data is distributed and stored in all the disk devices due to compression, the compressed data is read. It is stored in a RAID configuration. When compressed data does not reduce the read time as compared with the case where the uncompressed data is distributedly stored in all the disk devices, the data is stored in a duplicated structure.
Therefore, the read throughput does not decrease as compared with the case of the dual configuration. If a higher throughput can be obtained than in the case of using the duplex configuration, R
Since the data is stored in the AID configuration, the redundancy can be reduced and the data can be stored.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to embodiments.

FIG. 1 shows an outline of the configuration of a data storage system using a disk array device according to an embodiment of the present invention. The data storage system includes a host computer 11 that instructs reading and writing of data,
It is composed of first to fourth disk devices 12 to 15 for storing data in a distributed manner, and a disk array control device 16 for reading and writing data to these disk devices. The disk array control device 16 includes a host input / output control unit 17 that inputs / outputs data and various commands from / to the host computer 11. The compression processing unit 18 is a circuit device that compresses data to be written. The compressed data counting unit 19 counts the amount of data after being compressed by the compression processing unit 18.
The decompression processing unit 21 is a circuit that decompresses the compressed data read from each disk device. The parity generation unit 22 is a circuit unit that generates parity data for reconstructing data when the disk device is destroyed. The lower input / output control unit 23 is a circuit for performing an interface between the disk array control device 16 and each of the disk devices 12 to 15. The management unit 24 is a unit that performs overall control of these circuit devices according to a command from the host computer 11.

FIG. 2 shows an outline of the circuit configuration of the disk array controller shown in FIG. The disk array controller 16 plays a central role in its control.
A PU (Central Control Unit) 31 is provided. Various circuit devices are connected to the CPU 31 through various buses 32 such as a data bus. Of these, ROM (read only
A memory 33 is a memory for storing programs and fixed data. The RAM 34 is a random access memory for temporarily storing various data necessary for executing the program. First to fourth SCSI
(Small Computer Sysytem Interface) Controller 3
Reference numerals 5 to 38 are circuit devices that interface with the disk device, and the fifth SCSI controller 39 interfaces with the host computer 11. The compression processing circuit 41 is a circuit for compressing data to be stored in the disk devices 12-14. Compressed data counter circuit 42
Is a circuit for counting the data amount of the data after being compressed by the compression processing circuit 41. This circuit is designed to check whether the amount of data after being compressed has decreased. The decompression processing circuit 43 is a circuit portion that decompresses compressed data. The parity generation circuit 44 is a circuit that generates an error detection code for compressed data.

The first to fifth SCSI controllers 35 to 35
39, the compression processing circuit 41, the expansion processing circuit 43, and the parity generation circuit 44 are connected by a local bus 45. By transferring the data by the local bus 45, the transfer time inside the disk array control device 16 is shortened. Also, the first to fourth SCS
The I controllers 35 to 38 and the parity generation circuit 44 are
It is connected by a dedicated bus 46 for transmitting the data for which the parity code is generated and the generated parity code.

FIG. 3 schematically shows the flow of data inside the disk array controller. First to first
The storage area of the fourth disk unit 12-15, the initial storage area 51 ₁ to 51 ₄ for storing in a distributed input data from the host output control unit 17, respectively are prepared. Also, the first to third disk devices 12 to 1
The 4, compressed storage area 52 ₁ to 52 ₃ for storing the compressed data, the fourth disk device 15,
Parity storage area 5 for storing parity data
3 are prepared respectively. The data input from the upper input / output control unit 17 is the first to fourth disk devices 12 to
It is adapted to be stored in the 15 initial storage area 51 ₁ to 51 _4. Each area is managed in blocks of 4 kilobytes. Initial storage area 51 ₁
To 51 data D ₀ after being divided into blocks in the ₄
~ D _N are distributed and stored.

[0025] When the data in the initial storage area 51 ₁ to 51 ₄ are stored, the host computer 11 so that the write completion notification is sent. As a result, the host computer 11 is released from the writing process. Thereafter, the compression processing unit 18 compresses the data in block units, and the compressed data is stored in the compressed storage areas 52 _{1 to} 52 _{1 of the first} to third disk devices 12 to 14.
It is supposed to be stored in ₃ . Disk device 1
Compressed data F ₀ stored in each block 2 to 15
Parity codes corresponding to ~ F ₈ are created by the parity generator 22 and stored in the parity storage area 53 of the fourth disk device 15. In this way, create parity data corresponding to the compressed data,
The RAID configuration allows data redundancy. In this way, a method of creating a redundant configuration in which the data after compression has a RAID configuration will be referred to as a compression RAID configuration. After the time of reading the data compressed data is read from the compressed storage area 52 ₁ to 52 ₃ of the first to third disk device which has been extended by the decompression processing unit 21, the host computer 11 through the upper input-output control unit 17 It is supposed to be output.

FIG. 4 shows the registered contents of the data management table for managing the data storage destination in the disk array device. Data management table 61
Is composed of a storage destination address 62 that represents the address of the area where the data is stored, a data length 63 that represents the length of the stored data, and an attribute 64 that represents whether the data is stored compressed. Has been done. The storage destination address 62 stores the storage destination address inside the disk array control device 16 in association with the logical address 65 representing the address as seen from the host computer 11. Storage destination address, at the time when the writing of data is completed in the initial storage area 51 ₁ to 51 ₄ shown in FIG. 3, the initial storage address 66 are registered a storage address in this area. Also, so that value to the compressed storage address 67 is the address of this area when the data of the initial storage area 51 ₁ to 51 ₄ has been stored in compressed storage area 52 ₁ to 52 ₃ is compressed are changed ing. Data length 63, the data length in the initial storage area 51 ₁ to 51 _4, that is, the data amount in the state before being compressed are registered. Attributes 64, which represents whether the data is compressed, in the initial storage area 51 ₁ to 51 steps is stored in the ₄ "raw", at step stored compressed in the compression storage area "compression RAID "is registered.

FIG. 5 shows the flow of foreground processing performed by the disk array controller when a data write command is received. CPU31 is the fifth SCS
It waits for a write command to be sent from the host computer 11 via the I controller 39 (step S101). When the write command is received (step S10)
1; Y), the subsequently sent data which is divided into blocks (step S102), the initial storage area 51 ₁ of the first through fourth disk unit 12-15
51 ₄ dispersed stored (step S103). CP
U31 registers the address of the initial storage area corresponding to the first area storing the data in the data management table (step S104). Then, the attribute is set to "raw" (step S105). Then, a write completion notice is sent to the host computer 11 (step S10).
6). The processing up to this point is performed as foreground processing. Thereafter, the process of compressing the data in the initial storage area and storing it in the compressed storage area is performed as a background process. Therefore, the CPU 31 activates the background process (step S107) and ends the foreground process (end).

As described above, since the write completion notice is sent when the data is stored in the initial storage area, the host computer 11 can be released from the writing process early. Further, even when a large amount of data is written, the writing completion notice is sent at the time when the data is stored in the initial storage area of the disk device, so there is no need to prepare a large capacity cache memory.

FIG. 6 shows when a data write command is received.
Of background processing performed by the disk array controller
It represents the flow. CPU31 is the initial storage area
A 51₁~ 51_FourData is read from the compression processing circuit
This is compressed by 41 (step S201).
The compressed data is stored in the first to third disk devices 12 to 14
Compressed storage area 52₁~ 52₃Store in a distributed manner
(Step S202). Taking the case shown in FIG. 3 as an example
And the initial storage area 51 of the first disk device 12 ₁To
Stored data "D₀"Is compressed and the first data
Compressed storage area 52 of disk device 12₁"F₀"age
Stored. Similarly, "D₁"Is" F₁",
"D₂"Is" F₂Stored as “. Next, CPU 31
Are stored in the first to third disk devices 12 to 14
Create parity data for the compressed data (
(Step S203).
Stored in the priority storage area 53 (step S20)
4).

The parity data is created as follows. For example, in FIG. 3, “F ₀ ”, “F ₁ ”, and “F
Each block of ₂ ”and“ P ₀ ”is defined in advance as a corresponding block. Similarly,“ F ₃ ”,“ F ₄ ”,
Correlate “F ₅ ” and “P ₃ ”. In this way, the blocks of each disk device are associated in advance. The head data of the associated blocks are read out one byte at a time from the first to third disk devices 12 to 14, and the exclusive OR between the corresponding bits of these data is obtained. The result is stored as 1-byte parity data at the beginning of the corresponding block of the fourth disk device. Similarly, the exclusive OR of the second bytes and the third bytes is calculated and stored in the second and third bytes of the fourth disk device. This is performed for all data in one block, and further for all blocks storing compressed data. In the example shown in FIG. 3, “F ₀ ” ˜
Parity data for “F ₂ ” is stored in “P ₀ ” of the fourth disk device 15. "F ₃ " ~
The "P _3" for _{"F 5", "F 6} " ~ the "F _8" are respectively stored in the "P _6". In this way, the compressed data is stored in the RAID configuration.

Returning to FIG. 6, the description will be continued. After storing the data in the compressed storage area, the CPU 31 rewrites the storage destination address shown in FIG. 4 from the initial storage address to the compressed storage address (step S205). Furthermore, the attribute is changed from "raw" to "compressed RAID" (step S2).
06). Finally, the data stored in the initial storage area is erased (step S207), and the process ends (END). A considerable amount of processing time is required to compress the data as described above, generate parity data, and store the parity data in the compressed RAID configuration. However, by performing this in the background processing, the time required for writing seen from the host computer 11 is shortened.

FIG. 7 shows the flow of processing performed by the disk array control device when a read command is received. The CPU 31 waits for a read command from the host computer 11 (step S30).
1). When a read command is received (step S301;
Y), referring to the data management table corresponding to the data to be read, whether the attribute of the data is "raw",
It is checked whether it is "compressed RAID" (step S3).
02). In the case of “raw” (step S303),
Data is read from the initial storage area (step S30
4). In the case of "compressed RAID" (step S30
2; N), data is read from the compression storage area (step S304), and the decompression processing circuit 43 decompresses the data (step S305). After sending these data to the host computer 11, a read completion notice is sent (step S306). In the background processing, when the processing of re-storing the data with the compressed RAID configuration is not completed yet, the data is read from the initial storage area. Thus, even during background processing,
The data can be read.

First Modification

In the first modification, whether to store the data stored in the initial storage area in the compressed RAID configuration or in the duplex configuration to store the data is selected based on the required reading time. . That is, the data is stored in the compressed RAID configuration only when the required reading time in the compressed RAID configuration does not become longer than the required reading time in the redundant configuration. As a result, the required throughput can be ensured without increasing the number of disk devices, and the amount of data that can be accumulated by compression can be increased. The circuit configuration of the disk array control device in the first modified example is the same as that shown in FIG. 2 and its explanation is omitted.

FIG. 8 schematically shows the flow of data in the disk array device of the first modification. The same parts as those in FIG. 3 are designated by the same reference numerals and the description thereof will be omitted as appropriate. The amount of data after being compressed by the compression processing unit 18 is counted by the compressed data counting unit 19. Based on the amount of data after compression, it is determined whether the data should be stored in the compressed RAID configuration or the duplex configuration. Compression RAI
When the D configuration, the data is stored in compressed storage area 52 ₁ to 52 ₃ in the same manner as when shown in FIG. 3, the parity data is stored in the parity storage area 53 of the fourth disk device 15 It On the other hand, in the case of storing in a duplicated configuration, the first duplicated storage areas 71 _{1 to} 7 ₁
1 _4, the second redundant storage area 72 ₁ to 72 _4, the data of the initial storage area 51 ₁ to 51 ₄ are adapted to be stored is copied without any compression. When copying, a disk device different from the disk device storing the copy source data is selected as the storage destination. Further, the data of each block unit stored in the first duplicated storage area is stored in the second disk of a different disk device.
It is designed to be stored in the redundant storage area of.

For example, the block data “D ₀ ” of the initial storage area 51 ₁ of the first disk device 12 is stored in the first duplicated storage area 71 ₂ of the _second disk device 13 and the block data of the third disk device 14 of the third disk device 14. Redundant storage area 72 ₃
It is stored in. Similarly, the block data “D ₁ ” in the initial storage area 52 ₂ of the second disk device 12 is the third
First redundant storage area 71 ₃ of the disk device 14 and second redundant storage area 72 of the fourth disk device 15 of
Stored in ₄ , respectively. By thus cyclically shifting the disk device and storing the data,
It is not necessary to prepare a buffer memory for temporarily storing the data read from the copy source disk device.

FIG. 9 shows the registered contents of the data management table used when data is stored by selecting the duplex configuration and the compressed RAID configuration. The data management table 81 includes fields for registering a storage destination address 82, a data length 83, an attribute 84, and a required reading time 85. In the storage destination address 82, in addition to the initial storage address 86 and the compressed storage address 87, one of the duplex storage addresses 88 for registering the storage addresses of the data in the first and second duplex storage areas is registered. Further, in the column of the attribute 84, in addition to “raw” and “compressed RAID”, “duplex configuration” indicating that the data is stored in a duplex configuration is registered. In the column of the required read time 85, the uncompressed read time 9 indicating the read time when reading the duplicated data is shown.
1 and a compressed RAID read time 92 that represents the read time when reading the compressed RAID-structured data are registered. Further, a comparison result 93 which is a result of comparing these times is also registered.

The duplicated data is distributed to the first to fourth disk devices 12 to 15. Therefore, since the data can be read in parallel from the four disk devices, the uncompressed read time 91 is the amount of data before compression, the average read time of the disk device per block, and the disk device that stores the data in a distributed manner. It is calculated based on the number of vehicles. On the other hand, in the case of the compressed RAID configuration, one disk device for storing the parity data is required. Therefore, the compressed data is distributed and stored in the remaining three disk devices. Therefore, the compressed RAID read time 92 is the read time when the compressed data is distributed to the three disk devices and read from them in parallel at the same time. This is calculated from the amount of data after compression, the average read time, and the number of remaining disk devices other than those for storing parity data. When the amount of data does not decrease to some extent due to compression, the required reading time becomes shorter with the duplex configuration.

FIG. 10 shows a flow of foreground processing performed when a disk array control device having a redundant configuration of data by selecting a duplex configuration and a compression RAID configuration receives a write command. The CPU 31 waits for a write command from the host computer 11 (step S401). When the write command is received (step S401; Y), the data sent subsequently is sequentially divided into blocks (step S40).
2). The divided data are distributed and stored in the first to fourth disk devices 12 to 15 (step S40).
3). Next, the CPU 31 obtains the uncompressed read time and registers it in the data management table 82 (step S40).
4). By the compression processing circuit 41, the initial storage area 51 ₁
Compressing the data stored in the to 51 ₄ counts the amount of data by compressing the data counter circuit 42 (step S405). At this time, the compressed data is discarded without being stored. This is because the actual storage in the disk device increases the time required for the host computer 11 to be released from the writing process.

Based on the amount of data after compression, the compression RAI
The compressed RAID read time, which is the read time when the data is rearranged in the D configuration, is calculated, and is calculated.
2 is registered (step S406). The CPU 31 compares the obtained uncompressed read time with the compressed RAID read time,
The comparison result is registered in the data management table 82 (step S407). After that, the storage destination address of the data in the initial storage area is registered (step S408),
The attribute is set to "raw" (step S409). Now that the foreground processing has been completed, a write completion notice is sent to the host computer 11 (step S41).
0). The CPU 31 activates the background process (step S411) and ends the foreground process (end).

FIG. 11 shows a flow of background processing performed when a disk array control device having a redundant configuration of data by selecting a duplex configuration and a compression RAID configuration receives a write command. In the background processing, according to the comparison result obtained in the foreground processing,
Data is stored in a redundant configuration. First, the CPU 31 determines, based on the comparison result of the data management tables, which of the redundant configuration and the compressed RAID configuration should store the data in the redundant configuration (step S5).
01). When the comparison result indicates that short compression RAID read time (step S501; N), reads the data from the initial storage area 51 ₁ to 51 ₄ compresses it (step S502). Then, the data after compression dispersed in compressed storage area 52 ₁ to 52 ₃ of the first to third disk device 12-14 stores (step S50
3). Parity data corresponding to the compressed data stored in these first to third disk devices is generated by the parity generation circuit 44 (step S504), and this is generated as the fourth data.
The data is stored in the parity storage area 53 of the disk device 15 (step S505). The storage destination address of the data management table 82 is changed to the storage address in the compressed storage area (step S506), and the attribute is set to "compressed RAID" (step S507).

When the comparison result of the data management table 82 indicates that the compressed RAID read time is not short (step S501; Y), the initial storage areas 51 _{1 to} 5 ₁
The data stored in the first ₄ are copied to the first redundant storage area 71 ₁ to 71 ₄ stores (step S50
8). As for the relationship between the copy source and the copy destination disk devices, the disk devices are cyclically shifted one by one as shown in FIG. Furthermore, the initial storage area 51 ₁ to 51 ₄
Data only two minutes by cyclically shifting the disk apparatus of the stores are copied to the second duplexing storage areas 72 ₁ to 72 ₄ (step S509). Next, the storage destination address of the data management table 82 is changed to the storage address in the first and second redundant storage areas (step S51).
0), change the attribute to "redundant configuration" (step S5)
11). After storing in the redundancy by either duplicated configuration or compressed RAID configuration data, to erase the data stored in the initial storage area 51 ₁ to 51 ₄ (step S512). Here, the existence is deleted on the management table without actually deleting the data.

FIG. 12 shows a flow of processing performed when a disk array control device having a redundant configuration of data by selecting a duplex configuration and a compression RAID configuration receives a read command. CPU 31 is the host computer 1
It waits for a read command to be sent from 1 (step S601). When the read command is received (step S6)
01; Y), the data management table corresponding to the data to be read is searched, and it is determined whether or not the attribute is "raw" (step S602). If the attribute is "raw", that is, if the data has not yet been stored in the background processing in the redundant configuration (step S).
602; Y), is sent from the initial storage area 51 ₁ to 51 ₄ to the host computer 11 reads the data (step S603).

If the attribute is not "raw" (step
S602; N), whether it is "compressed RAID"
It is determined (step S604). Of "compressed RAID"
In this case, it is the storage destination address of the data management table
Based on the compressed storage address, the first to third disk devices are
Compressed storage area 52 of units 12-14₁~ 52₃From the day
Data is read (step S605). And read
After decompressing the data, send it to the host computer 11.
(Step S606). The attribute is "compressed RAID"
If not (step S604; N), the redundant configuration
Data is stored. In this case,
First based on the duplicated storage address of the data management table
~ First duplicated storage of the fourth disk unit 12-15
Rear 71 ₁~ 71_FourRead data from (step S
607). Here, the first redundant storage area 71₁~ 7
1_FourData is being read from, but the second redundant storage
Area 72₁~ 72_FourMay be read from. either
Data from the area of the host computer 1
1 and then a read completion notification is sent (step
In step S608, the process ends (end).

FIG. 13 is a schematic diagram showing the flow of data in the reading process when the disk device fails and data is lost. The same parts as those in FIG. 8 are designated by the same reference numerals, and the description thereof will be omitted as appropriate. 2nd in this figure
3 shows a data read route in the case where the disk device 13 has failed and data cannot be read. For example, "D _0" shall read data ~ "D _7", which is assumed to be stored are duplexed configuration. At this time, the data portion other than "D ₀ " and "D ₄ " can be read from the first duplicated storage area of the first, third and fourth disk devices 12, 13, 14. The data parts of "D ₀ " and "D ₄ " are
The second redundant storage area 72 of the third disk device 14
_It is read from part ₃ .

When the data is stored in the compressed RAID configuration, it is necessary to reconstruct the data stored in the compressed storage area 52 ₂ of the second disk device 13. For example, when reconstructing the compressed data “F ₁ ” in the compressed storage area 52 ₂ , the data is reconstructed by taking the exclusive OR between the data of the blocks associated with this of another disk device. Configure. That is, the data “F ₀ ” of the block of the first disk device 12 and the data “F ₂ ” of the third disk device,
The parity generation unit 22 calculates the exclusive OR between the corresponding bits of the parity data “P ₀ ” of the fourth disk device 15, and the calculation result becomes reconstructed data. Then, the compressed data read from the first and third disk devices 12 and 14 and the compressed data reproduced by the parity generation unit 22 are expanded by the expansion processing unit 21 and sent to the host computer 11. Has become.

FIG. 14 shows a flow of a read process performed by the disk array control device when the disk device is out of order. The CPU 31 waits for a read command from the host computer 11 (step S701). When the read command is received (step S701; Y), it is checked whether the disk device storing the data to be read is normal (step S702). If there is no defective disk device (step S702; N), normal read processing is performed (step S703). If there is a disk device that has a read failure (step S702;
Y), it is checked whether there are two or more defective disk devices (step S704). If there are two or more defective disk devices (step S
704; Y), since the data can no longer be read normally except when both the duplicated structure and the duplicate data holding disk device are normal, a predetermined error code is sent (step S705). If there is only one defective disk device (step S704; N), it is checked whether the attribute of the data to be read is "unprocessed" (step S706). When the attribute is "raw" (step S706; Y), since the data is not stored in the redundant configuration, normal data cannot be read. Therefore, a predetermined error code is transmitted (step S705), and the process is terminated.

If the attribute is not "raw" (step S706; N), it is checked whether it is "compressed RAID" (step S707). In the case of the compressed RAID configuration (step S707; Y), data is reconfigured from a normal disk device other than the defective disk device (step S708). Then, the compressed data read from the normal disk device and the compressed data reconstructed by the parity generation circuit are expanded by the expansion processing circuit 43 (step S709) and sent to the host computer 11. If the attribute is not "compressed RAID" (step S707; N), the data is stored in a duplicated configuration, so the data is stored in another disk device that stores the same data as the defective disk device. The part is read (step S710). Further, the data is read from the redundant storage area of the normal disk device (step S711) and sent to the host computer 11. When normal data is sent to the host computer 11, a read completion notice is sent (step S712), and the process ends.

Second Modification

When data for one block in the initial storage area is compressed and stored in one block in the compressed storage area, an empty area is generated in each block.
Therefore, in the second modified example, the data is packed and arranged so that this free area does not occur, and the amount of data that can be stored in the compressed RAID configuration is increased. Since the circuit configuration is the same as that shown in FIG. 2, its description is omitted.

FIG. 15 shows how the disk array control device of the second modified example is filled with empty areas. Initial storage in each block area 101 ₁ to 101 ₅ is stored is _{"D 0" ~ "D 4} " as the data before being compressed, respectively. When the data for each block is compressed, the amount of data decreases, and the compressed data 10 represented by "C ₀ " to "C ₄ " in the figure
It becomes 2 _{1 to 10 25} . Each compressed data 102 _{1 to} 102
_A predetermined end code is attached to the end of ₄ , indicating that the rest is empty. The shaded areas in the figure are empty areas generated in each block by compression. The compressed data C _{0 to} C ₄ (102 _{1 to} 102 ₅ ) are
The end code is deleted from each, and they are sequentially arranged in different areas in order from C ₀ . Then, it becomes the continuous compressed data 103 to which the end code is added at the end. Data 104 ₁ and 104 ₂ for each block size from the beginning of the continuous compressed data 103 are stored in the compressed storage areas of the first to third disk devices.

FIG. 16 shows the flow of background processing performed when the disk array control device of the second modification writes data. This process is performed after the foreground process shown in FIG. CPU31
Reads the data stored in the initial storage areas of the first to fourth disk devices 12 to 15 in block units and compresses it by the compression processing circuit 41 (step S80).
1). The compressed data is distributed and stored in predetermined work areas of the first to third disk devices (step S8).
02). The CPU 31 fills the empty area portion generated in each block of the work area and rearranges it as continuous data in the work area of the fourth disk device 15 (step S803). The data in the work area of the fourth disk device is read in block units, and this is read out as compressed storage areas 52 _{1 to} 52 ₁ of the first to third disk devices 12 to 14.
_The data is distributed to ₃ and stored (step S804). Compressed storage areas 52 _{1 to} 52 of these first to third disk devices
Parity data corresponding to the data of ₃ is created by the parity generation circuit 44 (step S805). This is stored in the parity storage area 53 of the fourth disk device (step S806), and the process ends (end). Although the change of the storage destination address and the attribute of the data management table is omitted in FIG. 16, it is performed in the same manner as shown in the flowchart of FIG.

FIG. 17 schematically shows the data flow inside the disk array control device of the second modification. The same parts as those in FIG. 3 are designated by the same reference numerals and the description thereof will be omitted. Initial storage area 51 ₁ to 51 ₄
The data of each block is compressed by the compression processing unit 18 in block units. The compressed data is stored in the remaining three disk devices except the fourth disk device 15 which stores the parity data. At this time, the data after compression is stored in a disk device different from the disk device in which the data before compression is stored. For example,
The compressed data “C ₀ ” of the data “D ₀ ” in the initial storage area 51 ₁ of the first disk device 12 is stored in the work area 111 ₂ of the second disk device 13. Each block in the work area of each disk device has an empty area as shown in FIG. These compressed data are continuously stored in the work area 112 of the fourth disk device. Since the data is copied from the first to third disk devices to the fourth disk device,
A buffer memory for temporarily storing the read data is no longer necessary.

Work area 1 of the fourth disk device 15
12, the data stored continuously, for each one block, is stored dispersed in the first to third disk device compressed storage area 52 ₁ to 52 ₄ 12 to 14. Parity data corresponding to these data is created by the parity generation unit 22 and stored in the parity storage area 53 of the fourth disk device 15. Initial storage area 51 at this point _{1-51 4} data and the work area 111 ₁
The data stored in 111 ₃ and 112 are erased. In this way, data is packed and compressed RA
By storing the ID, the storage capacity of the disk array device can be increased.

Third Modification

In the disk array control device of the third modification, a plurality of compression processing circuits and decompression processing circuits having different compression methods are prepared, and the data is compressed and stored by the compression method having the highest compression rate. There is.

FIG. 18 shows an outline of the circuit configuration of the compression processing circuit portion in the disk array control device of the third modified example. The circuit configuration other than the compression processing circuit and the decompression processing circuit is the same as that shown in FIG. 2, and the description thereof is omitted. In FIG. 2, the data output from the fifth SCSI controller 39 is the local bus 4
It is adapted to be sent to the compression processing circuit through 5 lines. The run-length compression circuit 121 includes a first FIFO (Firs
The data before compression is input via the memory 122. The run-length compression circuit 121 is a circuit that outputs a code in which the number of bits in which the same value has continued is expressed by a binary code every time the input binary data changes. The MMR (extended modified read) compression circuit 123 has a second FIFO.
The data before compression is input through the memory 124, and the data before compression is input through the third FIFO memory 126 to the dither compression circuit 125.

The MMR compression circuit 123 is a compression circuit that compresses data using vertical and horizontal related examples, and is particularly suitable for compression of image data in which characters and the like are drawn. The dither compression circuit 125 is a compression circuit suitable for compressing image data that is pseudo-multi-tone expressed by a dither matrix. The outputs of these compression circuits 121, 123, 125 are connected to the local bus 45. In addition, a bit counter 12 that counts the amount of data after compression
7 are provided. The bit counter 127 counts the amount of data after compression, and the compression circuit with the highest compression rate is selected. The read / write control circuit 128 is a control circuit that controls the operation of each circuit portion of the compression processing circuit.

FIG. 19 shows an outline of the circuit configuration of the expansion processing circuit portion in the disk array control device of the third modified example. The decompression processing circuit includes a run length decompression circuit 131, an MMR decompression circuit 132, a dither decompression circuit 133, and a read / write control circuit for controlling the operations of these circuits.
And a write control circuit 134. Each of the three compression circuits 121, 123, 1 shown in FIG.
It is an expansion circuit corresponding to 25.

FIG. 20 schematically shows the data flow in the background processing of the disk array controller of the third modification. The same parts as those in FIG. 8 are designated by the same reference numerals, and the description thereof will be omitted as appropriate. The data stored in the initial storage area 51 ₁ to 51 ₄ by foreground process, the first to third compression circuit 141 to 143
Are compressed in parallel at the same time. The amount of data after being compressed is counted by the bit counter 144. The counting results are compared by the comparison circuit 145 so that the compression circuit having the highest compression rate is selected. Data compressed by the selected compression circuit are respectively stored in the compressed storage area 52 ₁ to 52 ₃ of the first to third disk device 12 to 14, the parity data is generated by the parity generator 22 corresponding thereto It has become so. The parity data created is the fourth
The data is stored in the parity storage area 53 of the disk device 15.

FIG. 21 shows the registered contents of the data management table in the disk array controller of the third modification. However, since the storage destination address, the data length, and the attribute portion are the same as those in the data management table shown in FIG. 9, their description is omitted. Since the data management table 151 includes three compression circuits,
A field 152 for registering the compression method is provided.
The minimum compressed RAID read time column 153 registers the required read time when the data is compressed and stored by the compression method with the highest compression rate.

FIG. 22 shows the flow of processing performed when the disk array control device of the third modified example selects the compression method. CPU 31 has an initial storage area 51
₁ simultaneously output to 51 ₄ instructions for compressing the data that is stored in the three compression circuits 121, 123, 125, to perform compression of the data (step S901).
When the compression process is completed, the count values of the bit counter 127 connected to the respective compression circuits are compared, and the compression circuit with the highest compression rate is selected (step S90).
2). The compression method of the selected compression circuit is registered in the data management table 151 (step S903). afterwards,
The required reading time when the data is stored by rearranging it in the compressed RAID configuration is calculated and registered as the minimum compressed RAID reading time 153 in the data management table 151 (step S904).

In this way, by selecting a compression method having a high compression rate from a plurality of compression methods, the amount of data can be efficiently reduced and the storage capacity that can be stored in the disk array device can be apparently increased. Also, if the amount of data decreases, the time required for reading can be shortened accordingly.
A disk array device with high throughput can be obtained.

In the above-described embodiment and the first to third modifications, four disk devices are connected, but the number of disk devices is not limited to this. Compression R
In the case of the AID configuration, at least three disk devices are required. Also, when storing data in a duplicated configuration, the data is copied from the initial storage area to the first and second duplicated storage areas and the data in the initial storage area is erased. It is also possible to copy this to only one of the first and second duplicate storage areas and not erase the initial storage area, and store the data in a duplicated configuration by the duplicate storage area and the initial storage area.

Further, in the first to third modifications, the required reading time is calculated at the same time when it is stored in the initial storage area. If the compression processing takes longer than the storage time in the disk device, the compression processing and the processing for calculating the required reading time may be all performed in the background processing. Further, when the data is stored by the redundant configuration, the disk device is cyclically shifted, but it is not always necessary to cyclically shift it as long as the copy source disk device and the copy destination disk device are different. Also, when storing data in a compressed RAID configuration, the intermediate buffer memory can be made smaller or omitted by making the disk device storing the data before compression different from the disk device storing the data after compression. can do.

Besides, although three compression circuits are prepared in the third modification, the number of compression circuits is not limited to this. Any compression method may be used.

[0067]

As described above, according to the first aspect of the invention, the write completion notice is sent at the time when the data to be written is distributed and stored in a plurality of disk devices, and then the data is redundantly configured. And store it again. As a result, the host device that has issued the write command can be released from the writing process in a short time.

According to the second aspect of the present invention, when the data to be written is distributed and stored in a plurality of disk devices, a write completion notice is sent, and then the data is made redundant and re-stored. There is. When a read command is received before the re-storing process is completed, the data is initially read from the dispersedly stored areas. As a result, the data can be read even during the process of re-storing the data in the redundant configuration.

Further, according to the third aspect of the invention, even if the number of disk devices read in parallel at the same time decreases by one, the data whose read time is not compressed is distributed and stored in all the disk devices due to compression. When it becomes shorter than the case, the compressed data is stored in the RAID configuration. Further, when compressed data does not decrease as compared with the case where uncompressed read time is stored in all disk devices even if compressed, the data is stored in a duplicated structure. As a result, the throughput at the time of reading the disk device is
It does not decrease as compared with the case of the duplex configuration. That is, it is only necessary to prepare the number of disk devices that can satisfy the throughput required in the case of the redundant configuration. Further, when higher throughput is obtained than in the case of using the duplex configuration, the RAID configuration is used to store data. Therefore, it is possible to reduce the redundancy and store the data, and it is possible to increase the effective amount of data that can be stored in the disk array device as compared with the case where all the data are stored in the redundant configuration.

[Brief description of drawings]

FIG. 1 is a block diagram showing an outline of a configuration of a data storage system using a disk array device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an outline of a circuit configuration of the disk array control device shown in FIG.

FIG. 3 is an explanatory diagram that schematically shows the flow of data inside the disk array control device.

FIG. 4 is an explanatory diagram showing registered contents of a data management table for managing a storage destination of data in the disk array device.

FIG. 5 is a flow chart showing a flow of foreground processing performed by the disk array control device when a data write command is received.

FIG. 6 is a flow chart showing the flow of background processing performed by the disk array control device when a data write command is received.

FIG. 7 is a flowchart showing the flow of processing performed by the disk array control device when a read command is received.

FIG. 8 is an explanatory diagram schematically showing a data flow in a disk array device of a first modified example.

FIG. 9 is an explanatory diagram showing registration contents of a data management table used when data is stored by selecting a duplex configuration and a compression RAID configuration.

FIG. 10 is a flowchart showing a flow of a foreground process performed when a disk array control device having a redundant configuration of data by selecting a duplex configuration and a compression RAID configuration receives a write command.

FIG. 11 is a flowchart showing a flow of background processing performed when a disk array control device having a redundant configuration of data by selecting a duplex configuration and a compression RAID configuration receives a write command.

FIG. 12 is a flow chart showing a flow of processing performed when a disk array control device that selects a duplex configuration and a compression RAID configuration and has a data redundancy configuration receives a read command.

FIG. 13 is an explanatory diagram schematically showing a data flow in a reading process when a disk device fails and data is lost.

FIG. 14 is a flowchart showing the flow of a read process performed by the disk array control device when the disk device is out of order.

FIG. 15 is an explanatory diagram showing how empty areas are filled in a disk array control device of a second modified example.

FIG. 16 is a flowchart showing the flow of background processing performed when the disk array control device of the second modified example writes data.

FIG. 17 is an explanatory diagram schematically showing the flow of data inside the disk array control device of the second modified example.

FIG. 18 is a block diagram showing an outline of a circuit configuration of a compression processing circuit in a disk array control device of a third modified example.

FIG. 19 is an explanatory diagram schematically showing a data flow in background processing in the disk array control device of the third modified example.

FIG. 20 is an explanatory diagram schematically showing a data flow in background processing in a disk array control device of a third modified example.

FIG. 21 is an explanatory diagram showing registered contents of a data management table created in the disk array control device of the third modified example.

FIG. 22 is a flowchart showing the flow of processing performed when the disk array control device of the third modified example selects a compression method.

FIG. 23 is an explanatory diagram showing a data flow in a disk device that reconstructs data damaged by parity data.

[Explanation of symbols]

11 ... Host computer, 12-15 ... Disk device, 17, 23 ... Input / output control unit, 18 ... Compression processing unit, 1
9 ... Compressed data counting unit, 21 ... Decompression processing unit, 22 ...
Parity generator, 31 ... CPU, 32 ... Bus, 33 ... R
OM, 34 ... RAM, 35-39 ... SISC controller, 41 ... Compression processing circuit, 42 ... Compressed data counting circuit, 43 ... Decompression processing circuit, 44 ... Parity generation circuit, 4
5 ... Local bus, 51 ... Initial storage area, 52 ... Compressed storage area, 53 ... Parity storage area, 71, 72
… Redundant storage area

Claims (3)

[Claims] 1. A plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and data input from the write data input means are distributed to the plurality of disk devices and stored. The distributed storage means, the writing completion notification sending means for sending a predetermined writing completion notification indicating the end of writing when the data storage by the distributed storage means is completed, and the writing completion notification sending means. A disk array device, comprising: a redundant configuration storage unit for storing the data stored by the distributed storage unit in a redundant configuration and re-storing the data in the plurality of disk devices after a write completion notice is sent. 2. A plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and data input from the write data input means are distributed to the plurality of disk devices and stored. The distributed storage means, the writing completion notification sending means for sending a predetermined writing completion notification indicating the end of writing when the data storage by the distributed storage means is completed, and the writing completion notification sending means. Redundant configuration storage means for making the data stored by the distributed storage means redundant and storing it again in the plurality of disk devices after a write completion notice is sent, and read command receiving means for receiving a data read command. , The data to be read at the time when the read command is received by the read command receiving means is redundantly configured by the redundant configuration storing means and re-reproduced. A disk array device comprising: a reading means for reading the data stored by the distributed storage means when the data is not stored. 3. A plurality of disk devices, write data input means for inputting data to be stored in these disk devices, and data input from the write data input means are distributed and stored in the disk devices. At this time, first read time indexing means for obtaining a read time required to read this data simultaneously and in parallel from these disk devices, and compression means for compressing the data input from the write data input means, When the data after being compressed by the compressing means is distributed and stored in the disk devices other than the predetermined one among the plurality of disk devices, it is possible to read the data after the compression in parallel from these disk devices at the same time. A second read time indexing means for determining a required read time, and a read time calculated by the first read time indexing means for the second read time. First distributed storage means for storing the data input from the write data input means in a distributed manner in the disk device when the read time is shorter than or equal to the read time obtained by the indexing means, and the write data input means Second distributed storage means for storing the data input from the first distributed storage means and the storage device in which the data is distributed by changing the storage device, and the read time obtained by the second read time indexing means. Third distributed storage means for storing the data after being compressed by the compression means when the read time is shorter than the read time obtained by the first read time indexing means, by distributing the data to disk devices other than the predetermined one. When any one of the disk devices in which the data is stored by the third distributed storage means becomes unreadable, the data stored therein is reconstructed. And an error correction code storage unit for storing the error correction code generated by the error correction code generation unit in the predetermined one disk device. Disk array device characterized by.