WO2006072991A1

WO2006072991A1 - Write back method in raid device

Info

Publication number: WO2006072991A1
Application number: PCT/JP2005/000094
Authority: WO
Inventors: Kazuhiko Ikeuchi; Mikio Ito; Hidejirou Daikokuya; Shinya Mochizuki; Hideo Takahashi; Yoshihito Konta; Yasutake Satou; Hiroaki Ochi; Tsukasa Makino; Norihide Kubota
Original assignee: Fujitsu Limited
Priority date: 2005-01-07
Filing date: 2005-01-07
Publication date: 2006-07-13

Abstract

There is provided a write method in a RAID device capable of increasing the processing speed during update even when a process for updating a part of data stripe is contained. A RAID control device (3) includes: a cache memory (4); an update amount acquisition unit (6) for updating data stored in the cache memory (4) before being updated into updated data (5) and extracting information concerning the update amount of the data; a write method selection unit (7) for selecting a method for performing write back processing according to the information concerning the update amount extracted by the update amount acquisition unit (6); and a write processing unit (8) for performing write back processing by using the method selected by the write method selection unit (7).

Description

Write-back method in RAID device

Technical field

The present invention relates to a write back method in a RAID control system of a disk array subsystem.

Background art

[0002] RAID (Redundant Array of Inaependent / Inexpensive Disk drivesj) is widely used as a means of storing large amounts of data handled by computer systems. In particular, RAID5 is a cost-effective and simple configuration that protects data. It is widely used because of the reason why it can be realized.

[0003] RAID5 uses a parity disk to maintain data redundancy, so it is necessary to update the parity disk data (parity data) when performing write-back processing. .

[0004] Here, the write-back process refers to a process of writing information that is no longer needed on the cache memory to the disk, for example, when the disk power is new and information is taken out to the cache memory in the RAID controller.

FIG. 1 shows an example of data in a cache memory provided in a RAID control apparatus when all data stripes are updated during a write-back process.

Host computer equality The update data sent to the RAID controller is striped by the RAID controller, and if there is data to be updated (data before update) in the cache memory, the data before update is updated by the update data. Updated.

[0006] Here, in the following description, one piece of data striped by the RAID controller is referred to as a "data stripe".

Data 1 shown in the figure is RAID5 data consisting of four disks (Disk # 0— # 3 shown in the figure), and is striped to # 0— # 3! /.

[0007] Further, data 1 shown in the figure shows a case where data stripes # 0- # 2 are all update data. Since all the data stripes to be updated (data stripe # 0— # 2) are in the cache memory, generate the normative data (data stripe # 3) without reading from disk # 0— # 3. Can do. The generated data 1 is stored in Disk # 0— # 3 (hereinafter, this writing method is referred to as “: Bandwidth Write method”).

FIG. 2 is a diagram illustrating an example of data in the cache memory provided in the RAI D control apparatus when a part of the data stripe is updated during the write-back process.

Data 2 shown in the figure also shows RAID5 data consisting of four disks (Disk # 0— # 3 shown in the figure), and is striped to # 0— # 3 as shown in FIG. RU

[0009] Further, data 2 shown in the figure shows a case where only data stripes # 0 and # 1 are update data.

In this case, since only part of the data in the stripe is in the cache memory, the pre-update data is read from Disk # 0, # 1, and # 3, and XOR is performed with the updated data (data stripes # 0 and # 1). Note data needs to be generated. Then, the update data and the generated parity data are stored in Disk # 0, # 1, and # 3, respectively (hereinafter, this write method is referred to as “Small Write method”).

Patent Document 1: Japanese Patent Laid-Open No. 2003-296038

[0010] As described above, at the time of data update processing in RAID5 or the like, especially at the time of partial update of the update data (write processing to the disk), the data before the update and the parity data are also read out by the data collector, After creating new parity data (hereinafter referred to as "update parity") according to the update data, it is necessary to write the update parity along with the update data to the disk

[0011] In other words, when updating some data stripes, it requires more processing to read data before updating the disk power than when updating all data stripes. There was a problem that the speed became slow.

Disclosure of the invention

[0012] The present invention has been made in view of the above-described problems, and the problem to be solved is a processing speed at the time of update even when a process of updating some data stripes is included. Providing a write-back method in a RAID device capable of improving The

[0013] In order to solve the above-mentioned problem, the write back method in the RAID controller according to the present invention provides an update location of post-update data stored in a cache memory updated by update data sent from the host computer. The detection result power and an update amount acquisition process for acquiring the update amount of the updated data, and a write method for the updated data according to the update amount extracted by the update amount acquisition process, A write method selection process selected from a read method and a Read & Bandwidth Write method, a data stripe of the updated data stored in the cache memory is read, and the data stripe is updated by the method selected in the write method selection process And a writing process.

[0014] According to the present invention, the update amount acquisition process extracts the update amount of the post-update data stored in the cache memory updated by the update data sent from the host computer, and the write method selection process Select the optimal Write method according to the update amount, the Small Write method that is advantageous when the update amount is small, and the Read & Bandwidth Write method that is advantageous when the update amount is large. It is possible to improve.

[0015] Further, the present invention detects the update location of the updated data stored in the cache memory updated by the update data sent also by the host computer, and updates the updated data from the detection result. The update amount acquisition process for acquiring the data and the write method for selecting the write method for the updated data according to the update amount extracted by the update amount acquisition process from the Sma 11 Write method and the Read & Bandwidth Write method RAID processing includes a selection process and a data write process for reading the data stripe of the updated data stored in the cache memory and updating the data stripe by the Write method selected in the write method selection process. It may be a program for a RAID controller characterized by being executed by a controller.

In addition, the present invention detects an update location of post-update data stored in the cache memory updated by update data that is also sent by the host computer, and the detection result power update amount of the post-update data The update amount acquisition unit for acquiring the update amount and the update amount acquisition unit Thus, the write method of the updated data extracted according to the extracted update amount is selected from a small write method and a read & bandwidth write method, and a write method selection unit that selects the updated data stored in the cache memory. A RAID control device comprising: a write processing unit that reads a data stripe and updates the data stripe by a Write method selected in the write method selection process.

[0017] Further, the present invention is a RAID device including a plurality of storage units for storing update data sent by a host computer, and the post-update data stored in the cache memory updated by the update data An update amount acquisition unit that acquires an update amount of the post-update data, and an update amount of the post-update data according to the update amount extracted by the update amount acquisition unit. A write method selection unit that selects a write method from a small write method and a read & bandwidth write method, and reads a data stripe of the updated data stored in the cache memory, and is selected by the write method selection process. A RAID device comprising: a write processing unit that stores data in the plurality of storage units by a write method.

Brief Description of Drawings

FIG. 1 is a diagram showing an example of data in a cache memory included in a RAID control apparatus when all data stripes are updated during a write-back process.

FIG. 2 is a diagram showing an example of data in a cache memory included in a RAID control apparatus when a part of data stripes is updated during a write-back process.

FIG. 3 is a diagram illustrating the principle of the present invention.

FIG. 4 is a diagram illustrating a configuration example of a RAID device according to the present embodiment.

FIG. 5 is a flowchart showing processing of the RAID controller 3 according to the embodiment.

FIG. 6 is a diagram illustrating an example of update data management information according to the embodiment.

FIG. 7 is a diagram illustrating an example of update segment management information according to the embodiment.

FIG. 8 is a diagram showing an example of write method control information according to the present embodiment.

FIG. 9 is a diagram showing the relationship between update data management information and update segment management information.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to FIGS. FIG. 3 shows a principle diagram of the present invention.

The RAID controller 3 shown in FIG. 3 updates the cache memory 4 that temporarily stores data and the update data (not shown) that is stored in the cache memory 4 that is updated by the update data 5 that is transmitted by the host computer or the like. The update amount acquisition unit 6 that detects the location and extracts information about the update amount of the data, and a storage device (hereinafter simply referred to as “disk”) such as an HDD (Hard Disk Drive) connected to the RAID controller 3 Write method selection unit 7 that selects the method to perform Write Back processing according to the information about the update amount extracted by update amount acquisition unit 6, and update data that performs write back processing by the method selected by write method selection unit 7 It has at least a write processor 8 that sends 9 to the disk.

Furthermore, the update amount acquisition unit 6 according to the present embodiment includes an update data management information generation unit 6a and an update segment management information generation unit 6b.

The cache memory 5 is a storage unit that temporarily stores data read out from the disk power in response to a request from the host computer or the like.

[0021] When the pre-update data of the update data 5 sent by the host computer or the like is in the cache memory ₅ , the update amount acquisition unit 6 detects the update location where the pre-update data is updated by the update data 5. , Extract information about update amount.

That is, the update data management information generation unit 6a manages the update location updated by the update data 5 to which pre-update data (not shown) stored in the cache memory 4 is also sent by the host computer or the like. The data management information is generated, and the update segment management information (information related to the update amount) is extracted (generated) by the update segment management information generation unit 6b.

[0023] The write method selection unit 7 selects the Small Write method and the Read depending on the information about the update amount.

& Select the writing method by Write Back method from Bandwidth Write method.

[0024] The write processing unit 8 is the write method selected by the write method selection unit 7 in response to a Write Back processing command from a cache memory control unit (not shown) provided in the RAID controller 3. Write Back processing is performed to generate an update parity corresponding to the update data 5 stored in the cache memory 5 and transmit update data + update parity 9 to the disk.

FIG. 4 is a diagram illustrating a configuration example of the RAID device according to the present embodiment.

The RAID device shown in the figure includes at least a CM3a of a RAID controller connected to a host computer 14 via a CA (Channel Adapter) 13a and a disk array 16a connected to the CM3a via a router 15a. I have.

[0026] In the present embodiment, the RAID controller CM3b connected to the host computer 14 via a CA (Channel Adapter) 13b, and the disk array 16b connected to the CM3b via the router 15b are further provided. In addition, CM3a and CM3b are connected via a router 17. The disk array 16a and the disk array 16b are connected to CM3a and CM3b via the router 15a and the router 15b!

[0027] Here, in the present embodiment, the power shown in the configuration example when the RAID control device (CM3a, CM3a) and the disk array (disk array 16a, disk array 16b) are respectively duplicated is not intended to be limited to this. Absent. Redundant or higher redundant configuration may be used, or a single configuration using only CM 3a and disk array 16a may be used.

[0028] The CA 13a is an interface between an I / O device (not shown) provided in the host computer and the CM 3a, and controls commands and data between the host computers CM 3a.

[0029] The CM 3a includes at least a microprocessor 10a, a cache memory 11a, and a RAM (Random Access Memory) 12a, and operates the microprocessor 1 Oa according to a predetermined program, whereby the update data management information generation unit 6a and the update An update amount acquisition unit 6, a write method selection unit 7, and a write processing unit 8 composed of the segment management information generation unit 6b can be realized.

[0030] The update data management information generating unit 6a detects an update location stored in the cache memory 1 la and updated by the update data sent from the host computer 14 via the pre-update data tag CA 13a, Update data management information 18 for managing the update location for each block is generated and stored in the RAM 12a.

[0031] Further, according to the update segment management information 19 generated by the update segment management information generation unit 6b, the write method selection unit 7 selects the write method, and the write processing unit 8 selects the write method. According to the selected writing method, the updated data stored in the cache memory 1 la is written to the disk array 16a (write back processing is performed).

[0032] The disk array 16a is a RAID disk drive composed of Disk # 0— # 3 magnetic disks.

However, since this embodiment has a RAID5 configuration, a RAID disk drive consisting of at least three magnetic disks should be used instead of limiting the number of magnetic disks constituting the disk array 16a. Bho.

The configurations of the CA 13a, CM 3a, router 15a, and disk array 16a described above are the same for the CA 13b, CM 3b, router 15b, and disk array 16b, and a description thereof is omitted.

[0034] CM3a and CM3b are connected via a router 17. For example, when update data is sent from the host computer 14 to CM3a via CA13a, the processor 10a stores the update data in the cache memory 11a. And sent to CM3b via router 17. In the CM 3b, the update data received by the processor 10b is stored in the cache memory l ib so that the same data is always stored in the cache memory 1 la in the CM 3a and the cache memory 1 lb in the CM 3b.

[0035] Through the above processing, CM3a and CM3b are duplexed. Therefore, the operations (processing) of the CA 13a, CM 3a, router 15a, and disk array 16a will be described below for the sake of simplicity.

FIG. 5 is a flowchart showing the processing of the RAID controller 3 according to the present embodiment.

When there is no write back processing request from the cache memory control unit, etc., the write back processing in step S501 is not executed, and the update data management information generation unit 6a constituting the CM 3a which is the RAID controller 3 is Based on the update data received from the computer 14, the update location of the pre-update data stored in the cache memory 11a is detected in block units, and the update data management information 18 shown in FIG. 6 is generated. Then, the generated update data management information 18 is stored at a predetermined address in the RAM 12a.

Here, FIG. 6 shows an example of the update data management information 18 generated by the update data management information generation unit 6a.

The update data management information 18 shown in the figure contains update data sent from the host computer 14. Data number, data stripe number constituting update data of each update data number, block number constituting data stripe of each data stripe number, block update flag indicating presence / absence of update of the block of each block number, It consists of

[0038] Each block update flag shown in the figure is composed of, for example, 1-bit data. When it is 1, it indicates that there is an update, and when it is 0, it indicates that there is no update. Therefore, it is stored as a bitmap table on the RAM 12a!

[0039] When a write back process is requested to the processor 10a from a cache memory control unit (not shown) provided in the CM 3a, the CM 3a starts the write back process (step S50).

1) Move the process to step S502.

[0040] In step S502, the CM 3a refers to the update data management information 18 stored in the cache memory 11a, and determines whether all the block update flags are 1 (updated).

[0041] When all the block update flags are 1, it indicates that the updated data stored in the cache memory 11a has been updated by the update data sent by the host computer 14 as well, so the processing is step S503. Migrate to Then, the write method control information 22 shown in FIG. 9 is generated, and the write method flag for the corresponding update data is set to the Bandwidth Write method.

[0042] If all the block update flags are not 1 (at least one of the block update flags includes 0), a part of the pre-update data stored in the cache memory 11a is stored in the host computer. Since it indicates that the update data has been sent, the process proceeds to step S504.

In step S504, the block update flag of the update data management information 18 is rounded to the segment unit, and the update segment management information 19 shown in FIG. 7 is generated. Here, a segment is a unit that combines 16 blocks into one. In other words, in the process of step S504, a process for generating one segment update flag from the block update flag of each 16 blocks in the update data management information 18 is performed.

Here, FIG. 7 shows an example of the update segment management information according to the present embodiment.

The update segment management information 19 shown in the figure is an update sent from the host computer 14. The data number, the data stripe number that constitutes the update data of each update data number, the block number that constitutes the data stripe of each data stripe number, and the block number

It consists of a segment segment number as a unit for every 16 blocks, and a segment update flag indicating whether or not the segment of each segment number is updated.

[0045] Each segment update flag shown in the figure is composed of, for example, 1-bit data, and when it is 1, it indicates that all 16 blocks constituting the segment have been updated. In the case of, it indicates that at least one of the 16 blocks constituting the segment has been updated.

Here, FIG. 8 shows the relationship between the update data management information 18 and the update segment management information 19. FIG. 8 shows the block update flag 20 of the update data management information 18 and the segment update of the update segment management information 19. Flag 21 is shown.

For example, in step S504, the update segment management information generation unit 6b realized by the processor 10a included in the CM 3a refers to the block update flag 20 from a predetermined address of the update data management information 18 stored in the RAM 12a. However, if the block update flags for each 16 block are all 1, the segment update flag for the given address in the update segment management information 19 is set to 1, and at least the block update flag for each 16 block is set to 1. If one or more of them is 0, the segment update flag is set to 0.

[0048] Then, the update segment management information generating unit 6b stores the update segment management information 19 generated by the above processing in the RAM 12a, and the process proceeds to step S505.

In step S505, the write method selection unit 7 refers to the predetermined address force segment update flag in the update segment management information 19 stored in the RAM 12a, counts the number of segment update flags, and performs the process in step S506. To migrate.

[0049] In step S506, the writing method selection unit 7 compares the number of segment update flags counted in the process of step S505 with the threshold value, and if greater than the threshold value, the process proceeds to step S507. Then, the write method control information 22 shown in FIG. 9 is generated, and the write method flag for the corresponding update data is set to the Read & Bandwidth method. [0050] In this embodiment, 1Z2 of the total segment of the entire update data is used as the threshold value. In other words, the update data consists of three data stripes, one data stripe is 128 blocks, and one segment is 16 blocks, so 3 * (128/16) * (1/2) = 12 segments is the threshold. Become.

[0051] However, for example, the value of 1 Z3 of the total segment of the entire update data may be used as the threshold value, and the threshold value may be appropriately determined as necessary.

Here, the Read & Bandwidth method refers to, for example, update data 2 shown in Fig. 2 by reading data from Disk # 2 and performing an XOR with the update data to generate update parity and update data and the generated update Parity is stored in Disk # 0, # 1 and # 3 respectively.

In this Read & Bandwidth method, an updated parity is generated by reading the disk array power that is not updated by the updated data sent from the host computer 14 and reading the disk array power.

[0053] Therefore, if there are many updated data stripes (segments), there will be relatively fewer unupdated data stripes, so there will be less processing to read this unupdated data stripe from the disk array, and Write Back processing The processing speed is improved.

[0054] In step S506, if the number of segment update flags counted in the process of step S505 is larger than the threshold value, the process proceeds to step S508, and the corresponding update of the write method control information 22 shown in Fig. 9 is performed. Set the write method flag for the data to the Small Write method.

[0055] As shown in Fig. 2, this Small Write method reads the pre-update data stripe of the data stripe updated by the update data sent from the host computer 14, and generates the update parity by reading the disk array power. To do.

[0056] Therefore, if the number of updated data stripes (segments) is small, the number of processes for reading the pre-update data stripe corresponding to the updated data stripe from the disk array decreases, and the processing speed of the Write Back process Will improve.

As shown in FIG. 9, the write method control information 22 is generated by the processes of steps S503, S507, and S508 described above. The write method control information 22 shown in the figure is composed of an update data number and a write method flag for the update data of the update data number. In this example, when the write method flag is 2, it means the Bandwidth Write method, when the write method flag is power, it means the Small Write method, and when the write method flag is 0, it means the Read & B and width method. Power Data that can be identified by the Bandwidth Write method, Small Write method, and Read & Bandwidth method are not limited to this.

[0058] When the processes of steps S503, S507, and S508 are performed, CM 3a moves the process to step S509. If there is another update data, the process proceeds to step S502, and the processes in steps S502 to S509 are repeated.

[0059] If there is no other update data in step S509, the process proceeds to step S510, and the write processing unit 8 writes back the corresponding update data according to the write method control information 22. Perform processing.

[0060] As described above, when the update amount is small according to the update amount of the post-update data stored in the cache memory 4, only the data stripe of the update portion and the parity before update are stored in the disk array cache. In addition, when Write Back processing is performed by the Small Write method that can read and generate an update parity, and the amount of update data is large, only the data stripe without update is read from the disk array and the update parity is calculated. By performing Write Back processing using the Read & Bandwidth method that can be generated, the number of read operations (access) to the disk array can be reduced, which can improve the speed of update (Write Back) processing. It becomes possible.

[0061] In addition, the update segment management information 19 is generated from the update data management information 18, and the size of the update amount is determined by the segment update flag, thereby facilitating the determination process. For example, in the case of the present embodiment, as shown in FIG. 7, the update amount is represented by 16 segment update flags for one data stripe, so the update amount is larger or smaller in units of 2 bytes (16 bits). Can be determined.

In the above description, in step S 504 shown in FIG. 5, the segment update flag is set to 0 when at least one of the block update flags of 16 blocks is 0. The purpose is not limited to this.

[0063] For example, the condition for setting the segment update flag to 0 may be changed as necessary, such as when two or more of the block update flags of each 16 block are 0.

In addition, in this embodiment, one block is 512 bytes, one segment is 16 blocks, and one data stripe is eight segments (128 blocks). Yeah.

[0064] For example, every time update data is sent from the host computer 14, the CM 3a stores the update data size in the RAM 12a as a history, and the history power is also a block that constitutes one segment according to the change in the update data size. The number may be changed. This makes it possible to efficiently perform update segment management information generation processing (block update flag rounding processing).

Further, the update data management information 18 is not limited to the one shown in FIG. 6, but blocks the location where the pre-update data stored in the cache memory 11a is updated by the update data sent from the host computer 14 It only needs to be specified in units.

[0066] In addition, the update segment management information 19 is not limited to the one shown in FIG. 7, and the pre-update data stored in the cache memory 11a is updated by the update data sent by the host computer 14 as well. If the location can be specified in segment units.

[0067] Further, the write method control information 22 may be information related to a write method (for example, a small write method) with respect to update data, which is not limited to the information shown in FIG.

[0068] As described above, according to the present invention, a write back method in a RAID device that can improve the processing speed at the time of updating even when the processing of updating some data stripes is included. Can be provided.

Claims

The scope of the claims

[1] Host computer power Update amount acquisition for detecting the update location of the post-update data stored in the cache memory updated by the update data sent, and acquiring the update amount of the post-update data Processing,

A write method selection process for selecting a write method of the updated data according to the update amount extracted by the update amount acquisition process from a small write method and a read & bandwidth write method;

A write process for reading the data stripe of the updated data stored in the cache memory and updating the data stripe by a method selected by the write method selection process;

A write-back method characterized by causing a RAID controller to perform the operation.

[2] The update amount acquisition process includes:

Update data management information generation processing for generating update data management information for managing the presence or absence of update in the post-update data for each block and storing the update data management information in an update data management information storage unit;

The update data management information is read from the update data management information storage unit, and update segment management information for managing the update unit in the data stripe is generated for each segment composed of a plurality of the blocks. Update segment management information generation processing stored in the update segment management information storage means;

With

The writing method selection process is:

The update segment management information read from the update segment management information storage means V counts the number of segments indicating the update unit, and the write method of the data stripe is set according to the number of segments. 2. The write back method according to claim 1, wherein the write back method is selected from a method and a read & bandwidth width method.

[3] The updated segment management information generation process determines that the segment has an updated portion when a block equal to or greater than a predetermined threshold among the plurality of blocks constituting the segment is an updated block. 3. The write back method according to claim 2, wherein

[4] The write method selection process counts the number of segments indicating the update unit from the update segment management information, selects a small write method when the number of segments is equal to or less than a predetermined threshold, and selects the segment 4. The write-back method according to claim 3, wherein the read & bandwidth write method is selected when the number of data is equal to or greater than a predetermined threshold value.

[5] Host computer power Update amount acquisition for detecting the update location of the post-update data stored in the cache memory updated by the update data sent, and acquiring the update amount of the post-update data Processing,

In accordance with the update amount extracted by the update amount acquisition process,

Write method selection process for selecting Write method from Small Write method and Read & Bandwidth Write method,

Read the data stripe of the updated data stored in the cache memory

A write process for updating the data stripe by the method selected in the write method selection process;

A program for a RAID controller characterized by having the RAID controller perform the operation.

[6] The update amount acquisition process includes:

With

The writing method selection process is:

The update segment management information read from the update segment management information storage means V counts the number of segments indicating the update unit, and the write method of the data stripe is set according to the number of segments. Method and Read & Bandwidth Writ 6. The program for a RAID controller according to claim 5, wherein the program is selected from e type.

[7] The updated segment management information generation process determines that the segment has an updated portion when a block having a predetermined threshold value or more is an updated block among a plurality of blocks constituting the segment. 7. The program for a RAID controller according to claim 6, wherein the program is a RAID controller.

[8] The write method selection process counts the number of segments indicating the update unit from the update segment management information, selects a small write method when the number of segments is equal to or less than a predetermined threshold, and selects the segment 8. The program for a RAID controller according to claim 7, wherein the Read & Bandwidth Write method is selected when the number of files is equal to or greater than a predetermined threshold.

[9] Host computer power Update amount acquisition for detecting the update location of the post-update data stored in the cache memory updated by the update data sent, and acquiring the update amount of the post-update data And

A write method selection unit that selects a write method of the updated data according to the update amount extracted by the update amount acquisition unit from a small write method and a read & bandwidth write method;

A data processing unit that reads the data stripe of the updated data stored in the cache memory, and updates the data stripe by a method selected in the writing method selection process;

A RAID controller characterized by comprising:

[10] The update amount acquisition unit includes:

An update data management information generating unit for generating update data management information for managing the presence or absence of update in the post-update data after update for each block;

An update segment management information generating unit that generates update segment management information for managing an update unit in the data stripe for each segment composed of a plurality of the blocks;

With The writing method selection unit

Count the number of segments indicating the update unit for the update segment management information

10. The RAID controller according to claim 9, wherein the data stripe writing method is selected from a small write method and a read & bandwidth write method according to the number of segments.

[11] The updated segment management information generation unit determines that the segment has an updated portion when a block having a predetermined threshold value or more is an updated block among a plurality of blocks constituting the segment. 11. The RAID controller according to claim 10, wherein

[12] The write method selection unit counts the number of segments indicating the update unit from the update segment management information, selects a small write method when the number of segments is equal to or less than a predetermined threshold, and 12. The RAID control device according to claim 11, wherein the Read & Bandwidth Write method is selected when the number of segments is equal to or greater than a predetermined threshold.

[13] A RAID device having a plurality of storage units for storing update data sent by a host computer,

An update amount acquisition unit that detects an update location of the post-update data stored in the cache memory updated by the update data, and obtains an update amount of the post-update data.

A write processing unit that reads a data stripe of the updated data stored in the cache memory and stores the data stripe in the plurality of storage units by a method selected in the write method selection process;

A RAID device comprising:

[14] The update amount acquisition unit includes:

An update data management information generating unit for generating update data management information for managing the presence / absence of update in the post-update data for each detection block;

The update data management information power The data for each segment composed of a plurality of the blocks. An update segment management information generation unit for generating update segment management information for managing an update unit in a single stripe;

With

The writing method selection unit

Counting the number of segments indicating the update unit for the update segment management information, and selecting the data stripe writing method from the Small Write method and the Read & Bandwidth Write method according to the number of segments The RAID device according to claim 13.

[15] The updated segment management information generation unit determines that the segment has an updated portion when a block equal to or greater than a predetermined threshold among the plurality of blocks constituting the segment is an updated block. 15. The RAID device according to claim 14, wherein the RAID device is characterized in that:

[16] The write method selection unit counts the number of segments indicating the update unit from the update segment management information, selects a small write method when the number of segments is equal to or less than a predetermined threshold, 16. The RAID device according to claim 15, wherein the Read & Bandwidth Write method is selected when the number of segments is equal to or greater than a predetermined threshold.