CN104166601A - Storage data backup method and device - Google Patents

Abstract

The present invention proposes a method for backing up stored data, which is applied to a RAID array. The method includes: creating a write cache for a target RAID array; wherein, the write cache is enabled when the target RAID array fails; when the target RAID When the array fails, write the data written into the target RAID array into the write cache when executing a write IO command, and read data from the write cache when executing a read IO command; when the target RAID array recovers Normally, the write cache is automatically triggered to write back the written data to the target RAID array, and the write cache is disabled when the write back is completed. The invention can reduce hardware cost and performance impact on business.

Description

Backup method and device for stored data

technical field

The invention relates to the field of storage, in particular to a backup method and device for storing data.

Background technique

When the RAID array fails, all logical resources on the array cannot be read or written, which will cause service interruption. For example, if two disks in the RAID5 array cannot be read or written, or one disk in the RAID0 array cannot be read or written, the RAID array will fail. However, in most cases, the hard disk that cannot be read or written is probably just a problem with the hard disk adapter board, or the SAS connection of the DEU box is loose, but the disk itself is not a problem. When maintenance personnel restore the RAID array by replacing the adapter board or fixing the SAS connection, it usually causes the temporary failure of the RAID array. cause severe data loss. For example, in surveillance applications, all surveillance video during the failure of the RAID array will be lost.

In order to solve the above-mentioned problems, the mirroring technology is generally used at present to cope with the failure of the RAID array. Specifically, a mirrored RAID array may be first created for the RAID array, and a mirrored logical resource may also be created for the logical resource. When writing data to the RAID array, the data is simultaneously written to two independent RAID arrays (or logical resources). If the primary RAID array fails, the service flow is transferred to the mirror RAID array.

However, in the above solution, since a RAID array or logical resource of the same size must be used to create a mirror, the cost of hard disk space is doubled; when writing business data, two copies need to be written at the same time, which will affect the business writing performance when the system business is busy , and will double the load of system resources such as the CPU, thereby affecting the performance of other read and write services.

Contents of the invention

In view of this, the present invention proposes a backup method for storing data, which is applied to a RAID array, and the method includes:

Create a write cache for the target RAID array; wherein, the write cache is enabled when the target RAID array fails;

When the target RAID array fails, write the data written into the target RAID array into the write cache when executing a write IO command, and read data from the write cache when executing a read IO command;

When the target RAID array returns to normal, the write cache is automatically triggered to write back the written data to the target RAID array, and the write cache is disabled when the write back is completed.

Preferably, the method also includes:

In the write-back process, if a write IO command needs to be executed, the data corresponding to the write IO command is written into the target RAID array;

In the write-back process, if a read IO command needs to be executed, data is read from the write cache or from the target RAID array according to the data information to be read.

Preferably, the storage space of the write cache is divided into a global header, a RAID header, a Block header and a Block data area;

The global header records the size of the write cache storage space and whether each RAID header has been allocated the first bitmap bitmap;

The RAID header is divided into several sub-areas according to the number of target RAID arrays that use the write cache for failure backup, each sub-area corresponds to a target RAID array, and records the target RAID array information; the target RAID array information includes The array number, Block size and available Block number of the target RAID array;

Described Block head, each Block head corresponds to a Block data block respectively, and record respectively the RAID head number to which the Block data block belongs, the lba address of the Block data block in the target RAID array and record whether the Block data block has been written The second bitmap bitmap of the input data;

The Block data area includes several Block data blocks, and the size of the Block data blocks is the Block size recorded in the RAID header.

Preferably, said creating a write cache for the target RAID array includes:

Allocating an unused RAID header for the target RAID array according to the first bitmap;

Record the target RAID array information and the number of available blocks allocated for the target RAID array into a corresponding RAID header.

Preferably, writing the storage data written into the target RAID array into the write cache includes:

Find the corresponding RAID header according to the target RAID array number in the received write IO request;

According to the lba address in the write IO request, the Block header is searched in the Block header corresponding to the RAID header;

If the corresponding Block header is found, write the data into the Block data block corresponding to the Block header;

If do not find corresponding Block header, then obtain an empty Block data block, write data in this empty Block data block, record the serial number of belonging RAID head in the corresponding Block header of this empty Block data block simultaneously, in described Record in the second bitmap bitmap that the Block data block corresponding to the Block header has written data, and record the lba address of the data in the target RAID array into the Block header;

Wherein when an empty data block is obtained, the number of available blocks recorded in the corresponding RAID header is updated.

Preferably, reading data from the write cache includes:

Find the corresponding RAID header according to the target RAID array number in the received read IO request;

According to the lba address in the read IO request, search the Block header in the Block header corresponding to the RAID header;

If the corresponding Block header is found, and the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has written data, then read the data;

If the corresponding Block header is not found, or the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has no written data, read the data from the target RAID array, and return the read Get the result.

The present invention also proposes a backup device for storing data, which is applied to a RAID array, and the device includes:

Create a module for creating a write cache for the target RAID array; wherein, the write cache is enabled when the target RAID array fails;

A read-write module, configured to write the data written into the target RAID array into the write cache when executing a write IO command when the target RAID array fails, and read data from the write cache when executing a read IO command read data;

The write-back module is configured to automatically trigger the write cache to write back the written data to the target RAID array when the target RAID array returns to normal, and disable the write cache when the write-back is completed.

Preferably, the write-back module is further used for:

In the write-back process, if a write IO command needs to be executed, the data corresponding to the write IO command is written into the target RAID array;

The read-write module is further used for:

In the write-back process, if a read IO command needs to be executed, data is read from the write cache or from the target RAID array according to the data information to be read.

Preferably, the storage space of the write cache is divided into a global header, a RAID header, a Block header and a Block data area;

The global header includes the size of the write cache storage space and the first bitmap bitmap that records whether each RAID header has been allocated;

The RAID header is divided into several sub-areas according to the number of target RAID arrays that use the write cache for failure backup, each sub-area corresponds to a target RAID array, and records the target RAID array information; the target RAID array information includes The array number, Block size and available Block number of the target RAID array;

Described Block head, each Block head corresponds to a Block data block respectively, and record respectively the RAID head number to which the Block data block belongs, the lba address of the Block data block in the target RAID array and record whether the Block data block has been written The second bitmap bitmap of the input data;

The Block data area includes several Block data blocks, and the size of the Block data blocks is the Block size recorded in the RAID header.

Preferably, the creation module is further used for:

Allocating an unused RAID header for the target RAID array according to the first bitmap;

Record the target RAID array information and the number of available blocks allocated for the target RAID array into a corresponding RAID header.

Preferably, the read-write module is further used for:

Find the corresponding RAID header according to the target RAID array number in the received write IO request;

According to the lba address in the write IO request, the Block header is searched in the Block header corresponding to the RAID header;

If the corresponding Block header is found, write the data into the Block data block corresponding to the Block header;

If do not find corresponding Block header, then obtain an empty Block data block, write data in this empty Block data block, record the serial number of belonging RAID head in the corresponding Block header of this empty Block data block simultaneously, in described Record in the second bitmap bitmap that the Block data block corresponding to the Block header has written data, and record the lba address of the data in the target RAID array into the Block header;

Wherein, when an empty data block is acquired, the number of available Blocks recorded in the corresponding RAID header is updated.

Preferably, the read-write module is further used for:

Find the corresponding RAID header according to the target RAID array number in the received read IO request;

According to the lba address in the read IO request, search the Block header in the Block header corresponding to the RAID header;

If the corresponding Block header is found, and the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has written data, then read the data;

If the corresponding Block header is not found, or the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has no written data, read the data from the target RAID array, and return the read Get the result.

The present invention uses a small amount of storage space to provide failure backups for multiple RAID arrays at the same time by creating a write cache, which reduces hardware costs; moreover, because the write cache is only enabled when the RAID array fails, and can be restored when the RAID array fails , only the data written during the failure period is written back, which significantly reduces the performance impact on the business.

Description of drawings

FIG. 1 is a schematic diagram of the processing of external read and write IOs by a RAID storage array in the prior art;

Fig. 2 is a kind of RAID array shown in an exemplary embodiment of the present invention to the processing schematic diagram of external reading and writing IO;

Fig. 3 is a flow chart of a method for backing up stored data shown in an exemplary embodiment of the present invention;

Fig. 4 is a storage space organization form of a write cache storage medium shown in an exemplary embodiment of the present invention;

Fig. 5 is a kind of RAID array shown in an exemplary embodiment of the present invention and carries out the processing procedure of data invalidation backup and data recovery through write cache;

Fig. 6 is a block diagram of a backup device for storing data showing an exemplary embodiment of the present invention.

Detailed ways

Please refer to FIG. 1 . FIG. 1 is a schematic diagram of processing external read and write IOs by a RAID array in the prior art. As shown in the figure, when the RAID array is reading and writing data, the read and write IO request is sent to the RAID module through the upper module, and the Raid request sub-module judges the state of the RAID array to which the IO request belongs. If the state is normal, the IO request Send it to the RAID IO sub-module for processing, write data to the back-end hard disk, or read data from the back-end hard disk; if the array fails, return to the upper module to fail to read or write data.

In the invention, an independent write cache can be created for the RAID array, and the write cache can be used as a sub-module in the RAID module of the RAID array to participate in the processing of the external read and write IO by the RAID array.

Specifically, please refer to FIG. 2 . FIG. 2 is a schematic diagram of processing external read and write IOs by a RAID array shown in this embodiment. As shown in the figure, after the write cache sub-module is introduced, the read-write IO request is sent to the RAID module through the upper module, and the Raid request sub-module judges the state of the RAID array to which the IO request belongs. If the state is normal, the IO request is still sent to Handle it to the RAID IO sub-module, write data to the back-end hard disk, or read data from the back-end hard disk; if the array fails, read and write IO requests are no longer sent to the RAID IO sub-module for processing, but are processed by the write cache sub-module Processing, data reading and writing are all performed in the write cache. When the array recovers, the write cache is automatically triggered to write back the written data.

In order to make the technical solution of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and examples.

Referring to Fig. 3, in an exemplary embodiment of the present invention, a backup method for storing data is proposed, which is applied to a RAID array, and the method performs the following steps:

Step 301, creating a write cache for the target RAID array; wherein, the write cache is enabled when the target RAID array fails;

Step 302, when the target RAID array fails, write the data written into the target RAID array into the write cache when executing a write IO command, and read data from the write cache when executing a read IO command ;

Step 303, when the target RAID array returns to normal, automatically trigger the write cache to write back the written data to the target RAID array, and disable the write cache when the write back is completed.

In this embodiment, the storage medium for the write cache may be one or more independent memories, one or more independent hard disks, or one or more independent RAID arrays. Wherein, no matter which storage medium is used for the write cache, the storage space is organized according to the structure shown in FIG. 4 .

As shown in Figure 4, in this embodiment, the storage medium of the write cache is divided into four areas, and each area records the following information respectively: a global header, one or more RAID headers, multiple Block headers, and multiple Block headers. Data area, each area only records the parameters that must be saved when the power is turned off.

The following is a detailed description of the functions of the above four areas and the recorded parameters.

Global header area, which records the size of the write cache storage space (Size), the first bitmap bitmap (raid_bitmap) used to record whether each RAID header has been allocated, the controller ID (contol_id) to which it belongs, and the data calibration of this area Check value (check_sum). Wherein, the size of the write cache storage space can be set according to the actual business and the time of failure backup when creating the write cache; for example, assuming that there are 50 channels of 1Mbps video recording on the RAID array, 24h failure backup needs to be provided, so A RAID write cache of about 528G needs to be created. Then a 4TB hard disk is used as a write cache, which can provide 24h failure backup for seven RAID arrays of this kind of business at the same time.

The first bitmap bitmap is used to record whether each RAID header has been allocated. When a RAID header has a value of 1 in the corresponding position of the bitmap bitmap, it means that the RAID header has been allocated; when it is 0, it means that the RAID The header has not been allocated.

The belonging controller ID is used to distinguish the controller controlling the write cache. For example, in a specific application, in order to improve reliability, the target RAID array is usually controlled by two controllers (commonly known as dual control), and the controller controlling the target RAID array can be distinguished through the controller ID. Of course, in a non-dual-control environment, this parameter is not necessarily required.

The data check value is used to restore the lost parameters through check calculation when some parameters recorded in this area are accidentally lost.

RAID header area, which is divided into several sub-areas according to the number of target RAID arrays that use write cache for invalidation backup, and each sub-area corresponds to a target RAID array; for example, suppose the write cache needs to be invalidated for two target RAID arrays backup, then the RAID header area can be equally divided into two sub-areas, and each sub-area corresponds to a target RAID area. Each sub-area in the RAID header area records the corresponding target RAID array information and the data check value of the area respectively; the target RAID array information usually includes the array number (uuid), sector size (sector_size) of the target RAID array, Strip size (strip_size), Block size (block_size) and available Block number (block_number); wherein, the sector size and the stripe size depend on the specific configuration of the target RAID array; Assigned by the administrator when caching.

The block header area is divided into several block headers according to the total number of blocks in the write cache, and each block header corresponds to a block data block; each block header respectively records the block header number of the block header in the block header area ( index, which means the block header is the first block in the block header area), the RAID head number to which the block header belongs (raid_index, which means the block header belongs to the first RAID in the RAID header area), the block data block The lba address (logical block address, logical block address) of the recorded data in the target RAID array and whether the second bitmap bitmap (block_bitmap) of recording the Block data block has written data.

Block data area, this area is divided into several Block data blocks according to the total number of Blocks in the write cache, and the size of the Block data block is the size of the Block recorded in the RAID header. The total number of blocks in the write cache can be calculated by dividing the actual space size of the block data area in the write cache by the block size of a single block data block.

Please refer to FIG. 5. FIG. 5 shows the processing flow of the RAID array performing data failure backup and data recovery through the write cache shown in this embodiment. The processing flow of the RAID array performing data failure backup and data recovery through the write cache is described below in conjunction with FIG. 4 A detailed description.

In this embodiment, the write cache does not perform invalidation backup for any RAID array in the initial state. After the RAID array is created, if the RAID array needs to be invalidated, continue to create the write cache corresponding to the RAID array.

Specifically, when creating a write cache for the RAID array, you can use the bitmap bitmap according to the RAID header in the global header to allocate an empty RAID header for the RAID array, and then record the data of the RAID array in the assigned RAID header. Array number, sector size, stripe size, block size and other information, at the same time allocate the number of available blocks for the RAID array, and record the allocated number of available blocks in the RAID header. Wherein, when the number of available Blocks is specifically allocated, it can be evenly allocated to each target RAID array according to the total number of Blocks in the write cache; of course, in specific implementation, it can also be based on the actual business of the target RAID array, Different target RAID arrays allocate different numbers of available blocks, which is not particularly limited in this embodiment.

In this embodiment, the created write cache is disabled by default when the RAID array is operating normally, and is automatically enabled when the RAID array fails. When the write cache is enabled, the external read and write IO requests for the failed RAID array are processed by the write cache.

The write cache first searches for the corresponding Block header according to the target RAID array number and lba address in the received write IO request; specifically, when the write cache receives the write IO request, it first searches for the corresponding Block header according to the target RAID array in the write IO request Find the corresponding RAID header by serial number, and then search for the Block header in the Block header corresponding to the RAID header according to the lba address in the write IO request after finding the RAID header. If it is the first write, the content recorded in the block header is empty, and the corresponding block header cannot be found, so you can directly obtain an empty block data block, and write the data written this time into the empty block data block. And record the numbering of the RAID head in the Block header corresponding to the empty Block data block (that is, the numbering of the RAID head that is found according to the target RAID array number in the read IO request), and the second bitmap bitmap Corresponding bit position setting is 1; Simultaneously, the lba address (that is, the lba address in this write IO request) of this data in the target RAID array is also recorded in the Block header corresponding to this empty Block data block, to facilitate the target RAID array When recovering from the failure state to the normal state, the written data can be written back to the target RAID array according to the lba address recorded in the Block header.

When searching for the Block header according to the target RAID array number and sector address in the write IO request, if the corresponding Block header is found, it indicates that data has been written to the Block data block corresponding to the Block header, so the The written data is written into the Block data block corresponding to the Block header.

It is worth noting that when data is written into the write cache, when an empty Block data block is obtained, the number of available Blocks recorded in the corresponding RAID header needs to be updated, and the number of available Blocks is reduced by 1.

Further, when the write cache receives the read IO request, first, according to the received target RAID array number and the lba address in the read IO request, the block header is searched in the write cache; specifically, when the write cache receives the read After the IO request, first search for the corresponding RAID header according to the target RAID array number in the read IO request, and then search for the Block in the Block header corresponding to the RAID header according to the lba address in the read IO request. head;

If the corresponding Block header is found, and the corresponding bit in the second bitmap bitmap in the Block header has been set to 1, indicating that the block data block corresponding to the block header has written data before, then directly read fetch the data;

Of course, if the corresponding Block header is not found, or the corresponding bit position in the second bitmap bitmap in the Block header is set to 0, it indicates that the data that needs to be read this time is not recorded in the write cache, so from the Read data from the target RAID array and return the read result. Since the target RAID array is still in an invalid state at this time, a read failure is directly returned.

In this embodiment, part of the storage space may also be reserved in the write cache as the competing use space of each target RAID array. For the competing use space, even if any RAID head uses up all the available blocks (that is, the number of available blocks recorded in the RAID header is 0), if there are empty Block data blocks in the competing using space, then the RAID head can still use the competing Use the empty Block data block in the space. As for the size of the competing space, the user can configure it according to the business. Reserving space for competition in the write cache can enable more backup space to be used when only one target RAID array fails, thereby maximizing the backup time of the failed backup and providing maintenance personnel with longer troubleshooting time.

In this embodiment, when the target RAID array fails to enable the write cache, the maintenance personnel can go to the site to troubleshoot the problem as soon as possible and restore the array. For example, if the hard disk adapter board is faulty, or the SAS cable of the DEU box is loose, there is no problem with the disk itself. Maintenance personnel can restore the array by replacing the adapter board or SAS cable.

When the failed target RAID array is recovered, the write cache is automatically triggered to write back the written data to the target RAID array, and after all the written data in the write cache is written back, the write cache is automatically disabled.

In this embodiment, when the write cache is writing back, since the Iba address of the corresponding Block data block in the target RAID array has been recorded in the Block header, the written data can be written according to the Iba address. The write sequence is written back to the target RAID array. For example, the lba address can be converted into a pba address (physics block address, physical block address), and then the written-back data can be written into the corresponding physical block according to the converted pba address. Wherein, for the specific implementation of converting the lba address into a pba address, please refer to the prior art, which will not be described in detail in the present invention.

In this embodiment, during data write-back, new write IOs will not be processed by the write cache, and newly written data will be directly written into the restored target RAID array. However, because some valid data on the target RAID array is still in the write cache during the write-back period of the write cache, the read IO request during the write-back period is still processed by the write cache first, and misses when reading data from the write cache The corresponding data is read from the target RAID array.

Of course, in specific implementation, the write cache can also be enabled when the target RAID array is normal, and the written data of the target RAID array can be synchronously written into the write cache, so that even if the failure of the target RAID array is caused by a hard disk failure, All local data in the target RAID array will not be lost. For example, for applications in the field of video surveillance, the video index data needs to be read before the video is written, so the write cache can be enabled when the RAID array is normal, and the index data can be backed up in advance, so that when the RAID array fails, it can also be read normally. Get video index data.

As can be seen from the above description, the present invention can use a small amount of storage space to provide failure backups for multiple RAID arrays at the same time by creating a write cache, which reduces hardware costs; and, since the write cache is enabled when the RAID array fails, it can It effectively distinguishes the data written during the failure of the array, and when the RAID array fails and recovers, only the data written during the failure is written back, thereby significantly reducing the performance impact on the business. This solution is especially suitable for data backup and recovery of RAID array failure caused by non-disk failure.

Referring to Fig. 6, the present invention also proposes a backup device 60 for storing data, which is applied to a RAID array, and the device 60 includes:

Creating module 601, configured to create a write cache for the target RAID array; wherein, the write cache is enabled when the target RAID array fails;

The read-write module 602 is configured to write the data written into the target RAID array into the write cache when executing the write IO command when the target RAID array fails, and write the data written into the target RAID array into the write cache when executing the read IO command. read data in;

The write-back module 603 is configured to automatically trigger the write cache to write back the written data to the target RAID array when the target RAID array returns to normal, and disable the write cache when the write-back is completed.

In this embodiment, the write-back module 603 is further used to:

In the write-back process, if a write IO command needs to be executed, the data corresponding to the write IO command is written into the target RAID array;

The read-write module 602 is further used for:

In the write-back process, if a read IO command needs to be executed, data is read from the write cache or from the target RAID array according to the data information to be read.

In this embodiment, the storage space of the write cache is divided into a global header, a RAID header, a Block header, and a Block data area;

The global header includes the size of the write cache storage space and the first bitmap bitmap that records whether each RAID header has been allocated;

The RAID header is divided into several sub-areas according to the number of target RAID arrays that use the write cache for failure backup, each sub-area corresponds to a target RAID array, and records the target RAID array information; the target RAID array information includes Array number, Block size and available Block number of the target RAID array;

Described Block head, each Block head corresponds to a Block data block respectively, and record respectively the RAID head number to which the Block data block belongs, the lba address of the Block data block in the target RAID array and record whether the Block data block has been written The second bitmap bitmap of the input data;

The Block data area includes several Block data blocks, and the size of the Block data blocks is the Block size recorded in the RAID header.

In this embodiment, the creating module 601 is further used to:

Allocating an unused RAID header for the target RAID array according to the first bitmap;

Record the target RAID array information and the number of available blocks allocated for the target RAID array into a corresponding RAID header.

In this embodiment, the read-write module 602 is further used for:

Find the corresponding RAID header according to the target RAID array number in the received write IO request;

According to the lba address in the write IO request, the Block header is searched in the Block header corresponding to the RAID header;

If the corresponding Block header is found, write the data into the Block data block corresponding to the Block header;

If do not find corresponding Block header, then obtain an empty Block data block, write data in this empty Block data block, record the serial number of belonging RAID head in the corresponding Block header of this empty Block data block simultaneously, in described Record in the second bitmap bitmap that the Block data block corresponding to the Block header has written data, and record the lba address of the data in the target RAID array into the Block header;

Wherein, when an empty data block is acquired, the number of available Blocks recorded in the corresponding RAID header is updated.

In this embodiment, the read-write module 602 is further used for:

Find the corresponding RAID header according to the target RAID array number in the received read IO request;

According to the lba address in the read IO request, search the Block header in the Block header corresponding to the RAID header;

If the corresponding Block header is found, and the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has written data, then read the data;

If the corresponding Block header is not found, or the second bitmap bitmap in the Block header records that the Block data block corresponding to the Block header has no written data, read the data from the target RAID array, and return the read Get the result.

Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description in the embodiment, and can also be changed and located in one or more devices different from the embodiment. The modules in the above embodiments can be combined into one module, and can also be further divided into multiple sub-modules. The numbers of the above embodiments of the invention are only for description, and do not represent the advantages and disadvantages of the embodiments.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the present invention. within the scope of protection.

Claims (12)

1. store a backup method for data, be applied to RAID array, it is characterized in that, described method comprises:

For target RAID Creating is write buffer memory; Wherein, described in, write to be buffered in when described target RAID array loses efficacy and enable;

When described target RAID array lost efficacy, described in when IO order is write in execution, the data that write described target RAID array being write, write buffer memory, when IO order is read in execution from described write buffer memory reading out data;

When described target RAID array recovers normal, described in automatically triggering, write buffer memory written data is written back to described target RAID array, and described in forbidding, write buffer memory when write-back completes.

2. the method for claim 1, is characterized in that, described method also comprises:

In described write-back process, if need to carry out, write IO order, this is write to data corresponding to IO order and write described target RAID array;

In described write-back process, if need carried out, read IO order, according to data message to be read, from described, write reading out data buffer memory or from described target RAID array.

3. the method for claim 1, is characterized in that, described in write buffer memory storage space be divided into overall situation head, RAID head, Block head and Block data field;

Described overall head, the size in buffer memory space and the bitmap bitmap whether each RAID head has been assigned with write in record;

Described RAID head, writes the lost efficacy target RAID array number of backup of buffer memory according to use and is divided into some subregions, the respectively corresponding target RAID array of every sub regions, and record this target RAID array information; Described target RAID array information comprises array numbering, Block size and the available Block number of described target RAID array;

Described Block head, the respectively corresponding Block data block of each Block head, and record respectively lba address in target RAID array of RAID head numbering under described Block data block, this Block data block and record whether the 2nd bitmap bitmap of written data of this Block data block;

Described Block data field, comprises several Block data blocks, and the size of described Block data block is the described Block size recording in described RAID head.

4. method as claimed in claim 3, is characterized in that, describedly writes buffer memory and comprises for target RAID Creating:

According to a described bitmap bitmap, for described target RAID array distributes a untapped RAID head;

Described target RAID array information and the available Block number of distributing for described target RAID array are recorded in corresponding RAID head.

5. method as claimed in claim 3, is characterized in that, describedly writes buffer memory described in the storage data that write described target RAID array are write and comprises:

According to the target RAID array numbering of writing in IO request of receiving, search corresponding RAID head;

According to the described lba address of writing in IO request, in the Block head of this RAID head correspondence, search Block head;

If found corresponding Block head, data are write in the Block data block of this Block head correspondence;

If do not find corresponding Block head, obtain an empty Block data block, data are write in this sky Block data block, the numbering of RAID head under simultaneously recording in Block head corresponding to this sky Block data block, in described the 2nd bitmap bitmap, record the Block data block written data of this Block head correspondence, and the lba address in target RAID array is recorded in this Block head by these data;

Wherein, when getting an empty data block, upgrade the described available Block number recording in corresponding RAID head.

6. method as claimed in claim 3, is characterized in that, describedly from described, write reading out data buffer memory and comprises:

According to the target RAID array numbering of reading in IO request of receiving, search corresponding RAID head;

According to described lba address of reading in IO request, in the Block head of this RAID head correspondence, search Block head;

If found corresponding Block head, and the Block data block that described the 2nd bitmap bitmap in this Block head records this Block head correspondence write data, reads this data;

If do not find corresponding Block head, or the Block data block data writing not that records this Block head correspondence of described the 2nd bitmap bitmap in this Block head, reading out data from described target RAID array, and return to reading result.

7. store a back-up device for data, be applied to RAID array, it is characterized in that, described device comprises:

Creation module, is used to target RAID Creating to write buffer memory; Wherein, described in, write to be buffered in when described target RAID array loses efficacy and enable;

Module for reading and writing, for when described target RAID array lost efficacy, writes buffer memory described in when IO order is write in execution, the data that write described target RAID array being write, when IO order is read in execution from described write buffer memory reading out data;

Write-back module, for when described target RAID array recovers normal, writes buffer memory described in automatically triggering written data is written back to described target RAID array, and described in forbidding, write buffer memory when write-back completes.

8. device as claimed in claim 7, is characterized in that, described write-back module is further used for:

In described write-back process, if need to carry out, write IO order, this is write to data corresponding to IO order and write described target RAID array;

Described module for reading and writing is further used for:

In described write-back process, if need carried out, read IO order, according to data message to be read, from described, write reading out data buffer memory or from described target RAID array.

9. device as claimed in claim 7, is characterized in that, described in write buffer memory storage space be divided into overall situation head, RAID head, Block head and Block data field;

Described overall head, comprises and writes the size in buffer memory space and record the bitmap bitmap whether each RAID head has been assigned with;

Described RAID head, writes the lost efficacy target RAID array number of backup of buffer memory according to use and is divided into some subregions, the respectively corresponding target RAID array of every sub regions, and record this target RAID array information; Described target RAID array information comprises array numbering, Block size and the available Block number of described target RAID array;

Described Block head, the respectively corresponding Block data block of each Block head, and record respectively lba address in target RAID array of RAID head numbering under described Block data block, this Block data block and record whether the 2nd bitmap bitmap of written data of this Block data block;

Described Block data field, comprises several Block data blocks, and the size of described Block data block is the described Block size recording in described RAID head.

10. device as claimed in claim 9, is characterized in that, described creation module is further used for:

According to a described bitmap bitmap, for described target RAID array distributes a untapped RAID head;

Described target RAID array information and the available Block number of distributing for described target RAID array are recorded in corresponding RAID head.

11. devices as claimed in claim 9, is characterized in that, described module for reading and writing is further used for:

According to the target RAID array numbering of writing in IO request of receiving, search corresponding RAID head;

According to the described lba address of writing in IO request, in the Block head of this RAID head correspondence, search Block head;

If found corresponding Block head, data are write in the Block data block of this Block head correspondence;

If do not find corresponding Block head, obtain an empty Block data block, data are write in this sky Block data block, the numbering of RAID head under simultaneously recording in Block head corresponding to this sky Block data block, in described the 2nd bitmap bitmap, record the Block data block written data of this Block head correspondence, and the lba address in target RAID array is recorded in this Block head by these data;

Wherein, when getting an empty data block, upgrade the described available Block number recording in corresponding RAID head.

12. devices as claimed in claim 9, is characterized in that, described module for reading and writing is further used for:

According to the target RAID array numbering of reading in IO request of receiving, search corresponding RAID head;

According to described lba address of reading in IO request, in the Block head of this RAID head correspondence, search Block head;

If found corresponding Block head, and the Block data block that described the 2nd bitmap bitmap in this Block head records this Block head correspondence write data, reads this data;

If do not find corresponding Block head, or the Block data block data writing not that records this Block head correspondence of described the 2nd bitmap bitmap in this Block head, reading out data from described target RAID array, and return to reading result.