KR100472207B1 - RAID control system for Sharing Logical Disk Units by Multiple RAID Controllers - Google Patents

Abstract

The present invention solves an I / O bottleneck in a distributed shared RAID system that distributes I / O processing for each logical disk device to multiple controllers, and improves performance through multiplexing and data corruption or loss that may occur due to data sharing. The present invention relates to a data sharing shared RAID control system through multiple RAID controllers that can be prevented.

The present invention relates to a RAID control system for distributing and sharing data through multiple RAID controllers for a plurality of hosts and logical disk devices, wherein each RAID controller divides all the data of the shared logical disk device into a predetermined unit, each of a predetermined amount of data. And as a lock server for its own shared data area.

Description

Data distributed sharing RAID control system through multiple RAID controllers {RAID control system for Sharing Logical Disk Units by Multiple RAID Controllers}

The present invention relates to a distributed shared RAID control system, and more particularly, to solve input and output bottlenecks in a distributed shared RAID system in which multiple RAID controllers share logical disk devices, and data corruption or loss that may occur due to data sharing. The present invention relates to a data distribution sharing RAID control system through multiple RAID controllers.

Redundant Array of Inexpensive Disks (RAID) systems are disk array systems that organize and manage multiple physical disks to act as a single large disk.

In general, such a RAID system is connected to a storage device by inserting a controller in the form of an HBA to a host, or a system structure having several independent RAID controllers.

FIG. 1 is a block diagram of a RAID system having a plurality of such independent RAID controllers 120, and separates a plurality of controllers managing the host 110 and the storage device 130, and each RAID controller ( 120 may be configured to connect multiple storage devices through a storage area network (SAN) instead of only one host 110.

Each RAID controller 120 forms a RAID system independent of each other, each controller is responsible for the management of the logical disk device assigned to it.

Therefore, in the conventional RAID system, when the host 110 requests intensively for a specific disk device 130, the input / output request for other controllers is relatively small while the input / output request only for the controller managing the disk device. Because of this problem, performance bottlenecks occur in the controller, and there is a problem in that the failure of the controller and the performance degradation of the processing of stored data are caused.

In addition, in order to solve this bottleneck, when the I / O processing for each logical disk device is distributed to multiple controllers, an appropriate shared data management method is provided to prevent data corruption and loss that may occur due to data sharing. Required.

Accordingly, an object of the present invention is to solve the above-described problems and necessities presented, and an object of the present invention is to solve an I / O bottleneck in a distributed shared RAID system that distributes I / O processing for each logical disk device to multiple controllers. In addition, the present invention provides a data distribution sharing RAID control system through multiple RAID controllers that can prevent data corruption or loss that may occur due to performance improvement and data sharing through multiplexing.

In order to achieve the object of the present invention, a data distribution sharing RAID control system through a multiple RAID controller, RAID control system for sharing data through multiple RAID controllers for a plurality of hosts and logical disk devices, each RAID controller Divides all the data of the shared logical disk device into a certain unit, shares a certain amount of data, and operates as a lock server for its own shared data area.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a block diagram of a distributed shared RAID system according to the present invention.

As shown in FIG. 2, the distributed shared RAID system according to the present invention has a structure in which multiple RAID controllers 220 share each logical disk device 230.

In addition, a fast channel speed switch 240 is used for the interface between the plurality of hosts 210 and the multiple RAID controller 220 and the interface between the plurality of logical disk devices 230 and the multiple RAID controller 220. In order to speed up block data transfer between the RAID controllers 220, a scalable coherence interface (SCI) switch 250 is used.

In such a distributed shared RAID system, even though the number of logical disk devices 230 that one controller needs to manage increases, the I / O channels are multiplexed through multiple controllers, so that each RAID controller 220 needs to handle a plurality of controllers. Can be distributed processing.

However, it is very important to maintain data integrity because data access from any controller is more likely to be corrupted.

In order to solve this problem, the present invention uses each RAID controller 220 as a lock server to allow each RAID controller 220 to manage access rights to its own management data.

That is, each RAID controller 220 is a lock server that manages lock information on its own management data, and plays a role in charge of access right to the corresponding data when an input / output request for its own management data arrives.

The operation of a lock server is a centralized lock management in which one lock server is responsible for all data, and a distributed method in which multiple lock servers are in charge of data assigned to each of them. lock management).

The central method, which is a conventional lock server operation method, has a problem that the lock server is a bottleneck because requests for requesting data access permission to one controller are concentrated.

However, the distributed lock server operation according to the present invention is proposed to solve the centralization problem of the conventional central method, and effectively reduces the bottleneck of the central method by distributing data management responsibility through multiple lock servers. Can be.

The multiple RAID controllers 220 sharing one logical disk device 230 divides all data into a predetermined unit and manages a certain amount of data. Copies of the same data can exist in multiple RAID controllers 220 at the same time and the information is stored as lock information managed by the lock server controller.

Meanwhile, FIG. 3 is a diagram illustrating a shared data management structure of each RAID controller 220 according to the present invention.

As shown in FIG. 3, each RAID controller 220 is divided into a read cache 221, a write cache 222, and a backup write cache 224 in consideration of input / output performance.

At this time, each RAID controller 220 defines a backup write cache 224 for the write cache 222 in a round-robin manner for controller error recovery.

That is, the controller 2 has a backup write cache of controller 1, and the controller 3 has a backup write cache of controller 2. The backup write cache of the last controller is managed by controller 1.

Also, each RAID controller 220 manages lock information 223 for data managed by the RAID controller 220.

The lock information 223 is information related to whether or not to cache the data and the location of the caching controller. The RAID controller 220, which is a lock server, transmits its management data from the host 210 or another RAID controller 220. When there is a read or write request, the lock information 223 is searched first to check whether the corresponding data is cached and the caching controller, and then the corresponding operation is performed.

4 is a diagram illustrating a data distribution management structure of the multiple RAID controller 220 for one logical disk device 230 according to the present invention.

As shown in FIG. 4, one logical disk device 230 is divided into a plurality of segments having a predetermined size, and the size may be defined by a user.

 Each segment is defined by a logical block address, and its data can be stored across multiple physical disk devices.

At this time, each RAID controller 220 shares and manages segments of one logical disk device 230 and manages lock information 223 for data stored in segments managed by the RAID controller 220.

In addition, when updating the data managed by each lock server, the data must be transmitted to the lock server so that the lock server can be stored in the write cache 222 of the lock server. At this time, the deletion request for previous data existing in the other controller must be preceded.

5 is a diagram illustrating a lock management hash table and lock information structure managed by each RAID controller 220.

As shown in FIG. 5, the lock information 223 is managed in one hash table for each logical disk device 230. As shown in FIG. A hash table consists of several hash entries, and one hash entry joins a doubly linked list of lock information. The hash entry to store lock information is determined by indexing the remaining value obtained by dividing the physical disk block number stored by the lock information by the total number of hash entries.

The number of lock information 223 used for one logical disk device 230 is determined by the size of the disk cache defined when the logical disk device 230 is created and the number of controllers used.

The lock information 223 may also include a physical disk device number and a physical disk block number where data is stored, a number of processes waiting to refer to the lock information, a bitmap indicating controllers storing a copy of the data, and corresponding data. A flag indicating that a write is in progress, composed of process-synchronous semaphores.

The controller bitmap indicates controllers that read the corresponding data from the lock server and store the data in the read cache 221. The controller bitmap is composed of as many bits as the number of controllers and is assigned a unique bit number for each controller. A bit of 1 for a particular controller means that the data exists in the read cache 221 of that controller.

The controller bitmap information is used when a request for writing the corresponding data is requested to delete the previous data stored in the read cache 221 to the controller storing the previous data.

The write flag is also used to block data access while a write request for the data is in progress.

In other words, when a write request comes in, first set this flag to 1 so that all subsequent read and write requests will wait on the lock information until the ongoing write is completed. Start processing from the I / O request.

If the lock information 223 is present but both the bitmap and the write flag are 0, it means that the corresponding data exists in the write cache 222.

6 is a flowchart illustrating a process of processing a read request transmitted to a lock server controller managing data.

When a read request is received from the host 210, the RAID controller 220, which is a lock server, first searches for a lock information hash table managed by the lock server and checks whether there is lock information 223 for the data (S601).

At this time, if the lock information 223 does not exist, the new lock information 223 is allocated, the information is stored, the share bit of the controller bitmap corresponding to the controller is changed to 1, and then connected to the hash entry. (S602, S603)

After a new cache buffer is allocated to the read cache 221 (S604), data is read from the disk and stored in the read cache 221 (S605).

If the lock information 223 exists, the lock server RAID controller 220 checks the write flag to determine whether the write is in progress.

If the write is in progress, it enters the standby mode until the write is completed (S607).

If writing is not in progress, the controller searches for the share bits of the controller bitmap and stores the corresponding data (S608).

At this time, if the value of the share bit is 0, it means that it is stored in its own write cache 222, so that its own write cache 222 is searched (S609).

In addition, if the share bit for the self is 1, it means that it is stored in its own read cache 221. Therefore, the own read cache 221 is searched (S610).

If the share bit corresponding to the controller 220 other than itself is 1, after allocating a new cache buffer to its read cache 221 (S611), the logical disk device number and the physical disk device of the data to be read. The data read request message including the number, the physical disk block number, and the cache buffer address where the data is to be stored is transmitted to the other RAID controller 220 through the SCI switch, and then waits for a response.

When receiving a response from the other controller 220, it is determined whether an error is made through the response message. No error means that data has been successfully stored in another read cache 221 from another controller. In this case, by changing the share bit for itself to 1, it tells that it is also stored in its read cache (221) (S613).

When the read request data is successfully confirmed by the above process, the RAID controller 220 which is a lock server transmits the read data to the host 210 (S614).

FIG. 7 is a flowchart illustrating a process in which another RAID controller 220 processes a read request for data of the lock server RAID controller 220.

First, the other RAID controller 220 determines the lock server RAID controller 220 that manages the read request data, and then allocates a new cache buffer to its read cache 221 (S701).

The lock information request message including the logical disk device number, the physical disk device number, the physical disk block number, and the cache buffer address where the data is to be stored is transmitted to the lock server RAID controller 220 for the data. Wait for a response (S702).

If a response message is received from the lock server RAID controller 220 (S703), the share controller number for the data is checked through the response message contents (S704).

When the share controller includes itself, the other RAID controller 220 searches for its own read cache 221 because it means that the corresponding data is stored in its read cache 221.

In addition, when the share controller is the lock server RAID controller 220, data stored in the read cache 221 of the lock server RAID controller 220 is read.

If the share controller number is -1, it means that the corresponding data is not stored in any controller. After reading the physical disk device directly and storing the data in the read cache 221 (S706), the lock server RAID controller ( Since the corresponding data is stored in the read cache 221 of FIG. 220, the lock information change message is transmitted to change the share bit of the data to 1 (S709).

If the share controller number is another controller, it means that the other controller is stored in the read cache 221 of the other controller. The other RAID controller 220 sends a data read request message containing data information to be read to the other controller. Wait for a response after transmitting (S707).

When the response message is received from another controller (S708), if there is no error after checking whether the error is transmitted, the lock information change message is transmitted to the lock server RAID controller 220 (S709), and the data is read from its cache (221). Store in

Then, the data stored in its read cache 221 is transmitted to the host 210 through the above process (S710).

8 is a flowchart illustrating a process of processing a lock information request message transferred from another controller to a lock server controller when the lock server controller and the controller receiving the read request are different.

Upon receiving the lock information request message from the other RAID controller 220, the lock server RAID controller 220 first searches for the lock information 223 using the physical disk device number and the physical disk number stored in the request message. (S801)

At this time, if the lock information 223 is present but writing is in progress, it waits until the writing is completed (S802).

If the write is not in progress, the search bit is searched for the share bit, and if the value is 0, it is stored in the write cache 222. Thus, the lock server RAID controller 220 searches for the write cache 222 to retrieve data. After that (S803), the data is first stored on the disk. (S804)

After setting the share bit for the other RAID controller 220 that has read request to 1 (S805), the corresponding data is copied to the buffer address stored in the lock information request message (S807).

If the share bit value is not 0 and the share bit for itself is 1, after retrieving data from its read cache 221 (S806), the data is copied to the buffer address stored in the lock information request message (S807).

Then, a response message in which its controller number is stored is written and transmitted to the other RAID controller 220 which has requested this. (S808, S809)

If the share bit corresponding to the other controller other than the self is 1, the number of the other controller is stored in the response message and transmitted to the requested RAID controller 220. (S810, S809)

If the lock information 223 does not exist, the lock server RAID controller 220 sets the share controller number to -1, creates a response message, and transmits the response message to the request RAID controller 220. S811, S809)

9 is a flowchart illustrating a process of processing a data read request message transferred from another RAID controller 220.

When a RAID controller 220 other than the lock server RAID controller 220 for the data receives a read request message from another RAID controller 220, the RAID controller 220 stores the physical disk device number and the physical disk stored in the request message. The read cache 221 is searched using the block number (S901).

After checking whether an error occurs in the read cache search process, if there is no error, that is, if the corresponding data is stored in the read cache 221, the data is copied to the buffer address stored in the message.

After the response message including the error occurrence information is written (S903), it is transmitted to the other RAID controller 220 which has requested this (S904).

10 is a flowchart illustrating a process of processing a write request for data managed by the user.

Upon receiving a write request from the host 210, the RAID controller 220 first searches for lock information 223 corresponding to the block using the physical disk device number and the physical disk block number. (S1001)

In this case, if the lock information 223 does not exist, it means that the corresponding data is not stored in the write cache 222. Therefore, new lock information 223 is allocated and the write flag is set to 1 (S1002). After that, the cache tag is allocated to the write cache 222 (S1004), and the write data received from the host 210 is stored (S1005).

If the corresponding lock information 223 exists, the RAID controller 220 first checks whether the write flag is 1 (S1006).

If the write flag is 1, wait until the write is completed. (S1007)

If the write flag is 0 and the share bits are all 0, the corresponding data exists in the write cache 222, and the RAID controller 220 changes the write flag of the lock information 223 to 1 once. (S1008) The write data received from the host 210 is overwritten by the write cache 222. (S1009)

In addition, if the share bits are not all zeros, a data deletion request message is transmitted to the controllers corresponding to 1 to delete the corresponding data existing in the read cache 221 (S1010).

After receiving the response message indicating that the corresponding data of the read cache 221 has been deleted from all controllers having the share bit of 1, the cache tag is allocated to the write cache (S1004) and the data received from the host 210 is received. Save (S1005)

When the data storage is completed as described above, the write completion response message is transmitted to the host 210 (S1011), the write flag is changed to 0 (S1012), and the processes waiting for the corresponding lock information 223 are woken up.

11 is a flowchart illustrating a process of processing a write request for data managed by another controller.

When a write request for data managed by another controller comes from the host 210, the RAID controller 220 first allocates a temporary buffer (S1101) and then stores the data received from the host 210 in the temporary buffer. S1102)

After checking the lock server RAID controller 220 managing the data, the server sends a data write request message to the lock server RAID controller 220 and waits for the response (S1103).

When the response from the lock server RAID controller 220 is received (S1104), the temporary buffer allocation is canceled (S1105), and a response indicating that writing is completed is transmitted to the host 210 (S1106).

12 is a flowchart illustrating a process of processing a data write request message transferred from another RAID controller 220 to the lock server RAID controller 220.

Upon receiving a write request message from another RAID controller 220, the lock server RAID controller 220 first searches for lock information 223 corresponding to the write request message. (S1201)

At this time, if the lock information 223 does not exist, the new lock information 223 is allocated to store the data related information, the write flag is set to 1, and then connected to the hash entry (S1202, S1203).

After allocating a cache buffer to store data from the write cache 222 (S1204), data is taken from the buffer address stored in the write request message and stored in the allocated cache buffer (S1205).

If the lock information 223 exists, the write flag is first checked (S1206).

If the write flag is 1, wait for the current write to complete. (S1207)

If the write flag is also 0 and the share bits are all 0, after setting the write flag to 1 (S1208), new data is overwritten with the existing data stored in the write cache 222 (S1209).

If one of the share bits is present, the data deletion request message is transmitted to the corresponding RAID controllers 220 (S1210) and when the deletion completion message is received from the controllers receiving the message, the data is stored in the corresponding write cache 222. (S1205)

When data writing is completed, the write flag is changed back to 0 (S1211), and the write completion response message is transmitted to the other RAID controller 220 (S1212).

13 is a flowchart of a destaging process of storing write cache 222 data on a disk according to the present invention.

This destaging process saves new data because the dirty block of the write cache 222 is stored above a certain level when a read is requested from another RAID controller 220 for the data stored in the write cache 222. In order to do this, existing data is written to the disk from the write cache 222.

To this end, the RAID controller 220 searches its write cache 222 (S1301) and checks whether the dirty block is above a certain level (S1302).

When the dirty block is above a certain level, the RAID controller 220 first searches for the corresponding lock information 223 using the physical disk device number and the physical disk block number of the existing data to be stored on the disk, and then stores the lock information 223. Set the share bit for itself to 1 (S1303).

Then, a cache buffer usable from the read cache is allocated (S1304).

If valid data is stored in the allocated cache buffer, the corresponding lock information 223 is first found and changed (S1305 and S3106).

The cache buffer of the allocated read cache 221 tag and the cache buffer to be destaged are exchanged with each other (S1307), and dirty data is stored on the disk (S1308).

In this way, the destaged data is changed from the write cache 222 to the read cache 221 and stored.

The shared data management structure of the present invention as described above is a static distributed data management policy (Static, Distributed Data Management Policy).

That is, instead of managing all data stored in one logical disk device 230 in one controller, each controller 220 allocates data to the corresponding segments by dividing the data into segments in units of segments. Only manage access rights. As such, each controller manages the total capacity of one logical disk device 230 by a predetermined unit so that the data requested from several hosts 210 is data managed by different controllers at the same time, one logical disk device 230 Because I / O channels accessible by multiplexing are multiplexed, I / O channel bottlenecks in a single controller structure can be reduced.

In an application environment in which read requests for the same data block from multiple hosts 210 are frequently generated by multiple controllers 220, the controllers must sequentially access the corresponding data of the read cache in a single controller structure. In a structure in which the 220 manages distributed data, since a copy of the same data may exist in multiple controllers, read requests received through the various controllers 220 may be read from each controller 220 until a write request arrives. 221 may be processed at the same time.

In addition, even in the case of storing and managing different data for each controller 220, even if a request for data not managed by the controller comes from the host 210, the cache data of another controller is used without directly reading from the physical disk device. Can be processed.

As described above, the data distribution sharing RAID control system through the multiple RAID controllers according to the present invention can share performance through multiple RAID controllers without fixing the controller managing logical disk devices, thereby improving performance through I / O channel multiplexing. Not only can this be done, it can effectively prevent data corruption and loss due to data sharing.

In addition, read performance of the same data block from multiple hosts can be handled by multiple controllers at the same time, thereby improving read performance.

In addition, depending on how to determine the size of data that is distributed and stored and managed by each controller, even if a large number of hosts need to read data with less local characteristics, they can simultaneously process from multiple controllers. It is possible.

What has been described above is just one embodiment for implementing a distributed data sharing RAID control system through a multi-RAID controller according to the present invention, the present invention is not limited to the above-described embodiment, in the following claims Without departing from the gist of the present invention, any person of ordinary skill in the art will have the technical spirit of the present invention to the extent that various modifications can be made.

1 is a block diagram of a conventional RAID system having several independent RAID controllers.

2 is a block diagram of a distributed shared RAID system according to the present invention.

3 illustrates a shared data management structure of each RAID controller according to the present invention.

4 is a data distribution management structure diagram of multiple RAID controllers for one logical disk device according to the present invention;

5 is a data structure diagram of lock information and a lock management hash table according to the present invention;

6 is a flowchart of a read request processing in the lock server according to the present invention;

7 is a flowchart illustrating a read request process for data managed by another controller according to the present invention.

8 is a flowchart illustrating a lock information request message requested from another controller to a lock server according to the present invention.

9 is a flowchart of processing a data read request message requested from another controller to the lock server according to the present invention.

10 is a flowchart of a write request processing in the lock server according to the present invention.

11 is a flow chart for processing a write request for data managed by another controller according to the present invention.

12 is a flowchart of processing a data write request message requested from another controller to the lock server according to the present invention.

13 is a flow chart of a destaging process for storing write cache data on a disk according to the present invention.

<Description of the symbols for the main parts of the drawings>

210: host 220: RAID controller

230: logical disk device 250: SCI switch

221: read cache 222: write cache

223: lock information 224: backup write cache

Claims (15)

In a data distributed shared RAID control system, Multiple RAID controllers for distributing and sharing data among multiple hosts and logical disk devices. Each RAID controller divides the data area of the shared logical disk device to handle I / O of each partitioned area. Distributed RAID control system characterized in that it comprises a multiple RAID controller operating as a lock server to manage the access rights of. The system of claim 1, wherein the RAID controller uses an SCI switch for block data transfer with another RAID controller. The RAID controller of claim 1, wherein the RAID controller divides and manages the read cache, the write cache, and the backup write cache in consideration of input / output performance, and each RAID controller defines a backup write cache in a round-robin manner for controller error recovery. A distributed data sharing RAID control system using a multiple RAID controller, characterized in that for managing the lock information for the data managed by itself. 2. The system of claim 1, wherein the RAID controller updates data only in the lock server RAID controller and the update data exists only in the write cache of the lock server RAID controller. 2. The RAID controller of claim 1, wherein the RAID controller manages the lock information in one hash table for each logical disk device, the hash table is composed of several hash entries, and one hash entry is composed of lock information. A distributed data RAID control system using multiple RAID controllers characterized in that the list is connected. 2. The RAID controller of claim 1, wherein the RAID controller stores, as its lock information, a physical disk device number and a physical disk block number storing the shared data, the number of processes waiting to refer to the corresponding lock information, and a copy of the corresponding data. A data sharing shared RAID control system using multiple RAID controllers, comprising a controller bitmap representing controllers, a flag indicating that writes to the data are in progress, and process synchronous semaphores. 7. The RAID controller of claim 6, wherein the RAID controller assigns a unique bit number for each controller to a controller bitmap, and sets the share bit for the controller to 1 when the data exists in the read cache of a specific controller. When a share bit is 1, the controller having a share bit of 1 is requested to delete the read cache data. 7. The RAID controller of claim 6, wherein the RAID controller sets the write flag to 1 once a write request is made and all subsequent reads and writes when a write request comes in order to block data access while a write request for the corresponding data of lock information is in progress. A distributed data sharing RAID control system with multiple RAID controllers, characterized in that the request waits on the corresponding lock information until the write in progress is completed. In the control method of the data distribution sharing RAID control system according to claim 1, When a read request for data managed by the host is received, checking whether the lock information for the data exists by searching the lock information hash table; If the lock information does not exist, allocating the new lock information to set the share bit for itself to 1 and reading data from the disk and storing the read bit in the read cache; If the corresponding lock information exists, checking the write flag to check whether the write is in progress, and if the write on the data is not in progress, searching for the share bit to identify the controller that read the corresponding data; if all of the share bits are 0, searching for and transmitting its write cache; And If the share bit for itself is 1, it reads and delivers its own read cache. If the share bit for other controllers is 1, it sends a read request message to the controller and receives the data and sends it to its read cache. Storing and changing the share bit for itself to 1 and transferring the data to the host; controlling a distributed data sharing RAID control system through a multiple RAID controller. In the control method of the data distribution sharing RAID control system according to claim 1, Determining a lock server controller for the data and allocating a new cache buffer to its read cache when a read request is made to the lock server RAID controller for data not managed by the self; Sending a lock information request message including information about the data and its cache buffer address to the lock server controller; Ascertaining the number of the share controller for the data through the contents of the response message from the lock server controller; searching for and transmitting its read cache when it is included in the share controller; If the share controller is a lock server controller, it reads the data stored in the read cache of the lock server. If the share controller is another controller, it sends a read request message containing data information to the other controller and receives the response message. Transmitting a lock information change message to the lock server controller and storing the data in a read cache thereof; when the share controller number is -1 and no corresponding data is stored in any controller, reading the physical disk device directly, storing the corresponding data in a read cache, and transmitting a lock information change message to the lock server controller; And And transmitting the data stored in its read cache to the requesting host through each of the above-described steps. In the control method of the data distribution sharing RAID control system according to claim 1, Retrieving the lock information through the lock information request message when receiving a lock information request message for data managed by the other RAID controller; If the lock information does not exist, setting a share controller number to -1 to create a response message and transmitting the response message to the other controller; If the lock information exists but the current write is in progress, waiting until the write is completed; If the write bit is not in progress and the share bit is 0, changing the share bit for the other controller to 1, searching for its own write cache, and transmitting the data to the buffer address of the other controller; If the write bit is not in progress and the share bit of the bit is 1, retrieving the data from its read cache and transmitting the data to the buffer address of the other controller; If the write is not in progress and the share bit for another controller is 1, the number of the other controller is stored in the response message, and the response message is transmitted to the other controller as requested. Control method of the control system. In the control method of the data distribution sharing RAID control system according to claim 1, Retrieving lock information for the data if there is a write request for data managed by the host; If the lock information does not exist, allocating new lock information, setting a write flag to 1, and storing write data from the host in a write cache; If the corresponding lock information exists, first checking the write flag, and if the write flag is 1, waiting until the write is completed; If the write flag is 0 and the share bits are all 0, overwriting write data from the host with the write flag changed to 1; If all share bits are non-zero, requesting deletion of the corresponding data in the read cache from the controllers corresponding to 1, and if the share bits are deleted, storing write data from the host in the write cache; And Changing the write flag of the lock information to 0 when storing of the write data is completed through each of the steps; and controlling the distributed data sharing RAID control system through the multiple RAID controllers. In the control method of the data distribution sharing RAID control system according to claim 1, Allocating a temporary buffer to store data from the host when there is a write request for data managed by another controller from the host; Identifying a lock server controller for the data and transmitting a data write request message to the lock server controller and waiting for a response; And If a response is received from the lock server controller, canceling the temporary buffer allocation and transmitting a write completion response to a host; controlling a distributed data sharing RAID control system through a multiple RAID controller . In the control method of the data distribution sharing RAID control system according to claim 1, Retrieving the lock information corresponding to the write request message when receiving a write request message for data managed by the other RAID controller; If the corresponding lock information does not exist, allocating new lock information to store data related information, setting the write flag to 1, and taking write data from the buffer address of the other controller and storing the write data in a buffer of the write cache; If the corresponding lock information exists, first checking the write flag, and if the write flag is 1, waiting until the current write is completed; If the write flag is 0 and the share bit is also all 0, setting the write flag to 1 and overwriting new data with the existing data stored in the write cache; If all share bits are non-zero, send a delete request message for the data to controllers corresponding to 1, and if the delete is completed, set the write flag to 1 and store write data from another controller in the write cache. ; And And changing the write flag back to 0 when data writing is completed through each of the above steps. In the control method of the data distribution sharing RAID control system according to claim 1, In destaging the write cache, checking the write cache to determine whether the dirty block is above a certain level; retrieving lock information corresponding to existing data and setting the share bit for itself of the lock information to 1 when the dirty block is above a certain level; Allocating a cache buffer usable for the read cache; Finding and changing the lock information if valid data is stored in the allocated read cache buffer; And Exchanging the cache buffer of the allocated read cache tag with the write cache buffer destaged, and storing dirty data to a disk; controlling data sharing shared RAID control system through multiple RAID controllers Way.