CN104714758A - Method for building array by adding mirror image structure to check-based RAID and read-write system - Google Patents

Abstract

The invention discloses an array construction method and a read-write system based on verifying that RAID is added to a mirror structure; the array construction method includes four steps of address layout, data layout, data access and data reconstruction; the read-write system It includes an I/O module, a mirror image management module, an address conversion module, and a verification-based RAID module; the present invention adds a mirror image structure on the basis of a verification RAID, which improves the I/O performance of the array; construction time; the novel address layout method puts the original data and the mirrored data as adjacent as possible, shortening the moving distance of the magnetic head; updating the verification information in the background, alleviating the write amplification problem based on the verification RAID, and improving the performance of the array Write performance; due to the existence of mirrored data, the fault tolerance of the array is improved, so the availability and reliability of the array are greatly improved.

Description

An array construction method and read-write system based on verifying RAID and adding mirror structure

technical field

The invention belongs to the technical field of data storage, and more specifically relates to an array construction method and a read-write system based on a verification RAID and adding a mirror structure.

Background technique

Redundant Arrays of Independent Disks (RAID) has been widely used in large-scale distributed parallel storage systems because of its performance and reliability. In terms of performance, RAID achieves high throughput through striping. In terms of reliability, RAID uses redundant data such as mirroring or checksums to improve reliability.

RAID can be divided into three types according to its redundant structure. First, there is no redundant structure. The representative array is RAID0. RAID0 uses data striping technology (Data stripping) to disperse data to all disks, so that disks can be operated in parallel, which improves the read and write speed. The remaining structure has the highest space utilization rate in the array structure, but it also leads to low reliability. If any disk in the array fails, the entire system will be destroyed. Second, based on parity RAID, the representative array is RAID5. RAID5 also uses data striping technology to store data in all disks in blocks, so that the disks can be operated in parallel. Unlike RAID0, Each stripe has a parity value, and all parity values are evenly distributed among all disks, which has the advantage of improving reliability. If any disk in the array fails, the system can calculate it based on the remaining disks. The data of the failed disk; due to the addition of the redundant structure, the space utilization rate has decreased, and its redundant structure requires a total capacity of one disk. At the same time, due to its redundant structure, it leads to the problem of write amplification, that is, a write operation , will generate four actual IO operations, including two reads and two writes; in the reconstruction mode, it is necessary to read the data of all remaining disks to calculate the data of the faulty disk; in the degraded mode, the request to access the faulty disk is decomposed into multiple I/O requests, which will result in long refactoring times, degraded mode, and low I/O performance in refactored mode. Third, mirror-based RAID. The representative array is RAID1. RAID1 is also called disk mirroring. It mirrors the data of one disk to another disk. Since the data is copied on different disks, I/O offloading can be realized. , so that access can be evenly distributed in the mirror, and if any disk in the array fails, the data will not be lost. The request to access the faulty disk can be directly converted into the request to access its mirror disk, and will not be decomposed into multiple I /O, and the data of the mirror disk corresponding to the faulty disk can be directly read during reconstruction, the reconstruction time is short, and the I/O performance is high, but its space utilization rate is poor, and only half of the array capacity can be used.

Contents of the invention

Aiming at the above defects or improvement needs of the prior art, the present invention provides an elastic data mirroring system and method for verification disk arrays, the purpose of which is to add a mirror structure on the basis of verification RAID to improve the fault tolerance of the array ability.

In order to achieve the above object, according to one aspect of the present invention, there is provided a method for constructing an array based on verifying RAID and adding a mirror structure, wherein the method for constructing an array includes address layout, data layout, data access and data reconstruction Four steps, as follows:

(1) Address layout:

(1.1) Set the number of stripes in the segment according to the number of disks M, and the number of stripes in the segment is an integer multiple of M;

(1.2) determine the segment number K according to the number of data blocks in the disk and the number of stripes in the segment, the number of segments K=the number of data blocks in the disk/segment number of stripes;

(1.3) Get N=K/2; Interleave the segments whose segment numbers are 1 to N with the segments whose segment numbers are N+1 to K: store the segment whose number is N+1 after the segment whose number is 1, and then store For the segment numbered 2, the segment numbered N+2 will be stored after the number 2, and stored interleaved in this way; if K/2 is a non-integer, it will be rounded;

(2) Data layout:

(2.1) The data in the original data segment is in the same layout as the parity-based RAID;

(2.2) On the basis of the layout of the original data segment, the mirrored data segment changes the disk number stored in the data block in the same stripe, and the disk number is uniformly shifted to the right or left; uniformly shifted to the right, that is, the disk number plus i, If the disk number after adding i is greater than M, then subtract M; uniformly shift to the left, that is, the disk number minus i, if the disk number after subtracting i is less than 0, then add M;

(3) Data reading and writing:

(3.1) Calculate the physical address of the data to be accessed; if it is a write request, also calculate the physical address of the verification information;

(3.2) Judging whether the data to be accessed has a mirror image; if there is no mirror image, then enter step (3.3), if there is a mirror image, then enter step (3.4);

(3.3) Send a read and write request to the disk pointed to by the physical address of the data; if it is a write request, update the verification information of the original data;

(3.4) Calculate the physical address of the mirrored data; if it is a write request, also calculate the physical address of the mirrored data verification data, send the write request to the disk pointed to by the corresponding data physical address and the disk pointed to by the mirrored physical address, and update The verification information of the original data and the mirrored data; if it is read, balance the load size of the disk where the original data is located and the disk where the mirrored data is located, and select the one with the smaller load to issue a read request;

(4) Data reconstruction:

If there is a faulty disk, determine whether the data on the faulty disk has a mirror image. If there is a mirror image, read the corresponding data from the mirror image and write it to the backup disk; Other data blocks calculate the data of the faulty disk and write it into the backup disk.

Preferably, when writing data, if the array space is insufficient, the mirrored data space is elastically released according to the priority specified by the user. When the capacity is half of the capacity, all data has mirror redundancy; when the user data needs more than half of the array capacity according to the parity-based RAID organization, the user can specify the priority of the data and release the data with low priority first segment mirroring; as user data increments to saturation, the array degenerates into a parity-based RAID.

Preferably, the method for judging whether there is a mirror image is specifically: judging according to the content recorded in the strip segment bitmap in the mirror image management module, each segment uses 2 bits to record the state, and "00" means idle; "01" means that there is data There is no mirror image; "10" means that there is data and its mirror data is in the downward adjacent position; "11" means that there is data and the data has a mirror image, and the mirror position needs to be searched through the mirror mapping table.

Preferably, there are two ways to update the verification information: first, if the user does not require the verification data to be updated synchronously, then use the verification delay queue in the I/O module for background processing; second, synchronous processing, in When writing original data and mirrored data, the verification information of the original data and the verification information of the mirrored data are updated synchronously.

Preferably, in the absence of mirroring, during the execution of the read request, if the disk where the original data is located fails, the read is performed according to the method of verifying the RAID; this method of following the verification of the RAID read operation is degraded read .

Preferably, in the case of mirroring, during the execution of the read request, if the disk where the original data is located fails, the user read request is directly sent to the disk where the mirrored data is located.

In order to realize the purpose of the present invention, according to another aspect of the present invention, a kind of array reading and writing system based on verification RAID adding mirror structure is provided, it is characterized in that, described system comprises I/O module, mirror image management module, address Transformation module, RAID module based on verification;

The I/O module receives the upper layer read and write request, and outputs the logical address information included in the read and write request; the mirror management module judges and searches for the mirror data of the original data pointed to by the address according to the logic address information, and outputs the mirror data logic Address; the address translation module receives the logical address of the original data and the image data, and outputs the physical address of the original data and the image data; if it is a write request, it also outputs the physical address of the original data and the image data inspection block; the I/O module is based on the described The physical address sends read and write requests to the corresponding disks; the RAID module based on parity is used to execute read requests when the disk where the original data is located fails without mirroring.

Preferably, the I/O module includes a check delay queue for updating the check information of the data segment in the background.

Further preferably, the verification delay queue can be extended by 8 bits or 16 bits; the extended queue is used to record the segment number of the data block that needs to be updated and the corresponding offset stripe number, so as to realize small Granular updates.

Preferably, the mirror management module includes a stripe segment bitmap and a mirror mapping table, the stripe segment bitmap records the state of the stripe segment, and the mirror mapping table records the positions of non-adjacent mirror images; by traversing the stripe segment bitmap Find free segments and allocate them to original data and mirrored data; if the space allocated for mirrored data is not adjacent downward, record the location of non-adjacent mirrors in the mirror mapping table; the mirror management module is used to manage space and mirror data : Allocate space for the original data, allocate the image; find the corresponding image according to the original data, and elastically release the image data space according to the priority and array capacity specified by the user when the space is insufficient.

Generally speaking, compared with the prior art, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

(1) Since the mirror structure is added based on the verification RAID, the mirror image of the data is stored in the free space. During the execution of the read request, the mirror image can be used to balance the disks, and the mirror image can be returned directly when the read is downgraded. Data; during the execution of the write request, the verification information can be delayed to be updated, alleviating the write amplification problem based on the verification RAID, and improving the read and write performance of the array;

(2) Due to the addition of the mirror structure, the mirror data is used for data reconstruction, which shortens the time for data reconstruction;

(3) Since the present invention has adopted the address layout mode of interleaved storage, the original data and the image data are placed in adjacent positions as much as possible, shortening the moving distance of the magnetic head, and improving the read-write speed of the array;

(4) Since the I/O module of the present invention introduces a delay queue, the background update of the verification information is realized by using the delay queue, which alleviates the write amplification problem based on the verification RAID, and improves the write performance of the array;

(5) Due to the addition of the mirror structure and the existence of mirror data, the fault tolerance of the array can reach 1 to M/2 disks (M is the total number of disks in the array), which improves the availability and reliability of the array and improves the system performance. performance.

Description of drawings

The block diagram of the array reading and writing system based on the verification RAID added to the mirror structure provided by Fig. 1 embodiment 1 of the present invention;

FIG. 2 is a schematic diagram of the data structure provided by Embodiment 1 of the present invention;

Fig. 3 is a schematic diagram of RAID5 data layout in Embodiment 2 of the present invention;

FIG. 4 is a schematic diagram of the data layout applied to RAID5 by the array construction method based on verifying that RAID is added to the mirror structure provided by Embodiment 2 of the present invention;

Fig. 5 is a schematic diagram of the segment address layout in Embodiment 2 of the present invention;

6 is a flow chart of a data writing method in Embodiment 2 of the present invention;

Fig. 7 is a flow chart of the data reading method in Embodiment 2 of the present invention;

Fig. 8 is a flow chart of the data reconstruction method in Embodiment 2 of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other.

FIG. 1 shows an array read-write system based on verifying RAID and adding a mirror structure provided by Embodiment 1 of the present invention. For convenience of description, only parts related to the present invention are shown.

The system provided by embodiment 1 includes an I/O module, a mirror image management module, an address translation module and a RAID module based on verification;

Described I/O module accepts upper layer read and write request; Mirror image management module judges the image data segment of the original data segment that search address points to according to the address information in the read and write request, returns mirror image data segment address; Address conversion module logical address according to Segment address layout and data layout based on verification RAID are mapped to physical addresses on the disk. If it is a write request, it is also necessary to calculate the physical address of the verification block corresponding to the original data and mirror data; the I/O module decomposes the read and write requests, and the following Send to the disk corresponding to the physical address; based on the verification RAID module, it is used to execute the read request when the disk where the original data is located fails without mirroring.

The I/O module of embodiment 1 has verification delay queue, as shown in verification delay queue 11 among Fig. 2, is used for the verification information of background update data segment; In writing process, if user does not require verification data For synchronous update, the segment number will be recorded in the verification delay queue. When the system is idle or the load is not high, the verification information of the recorded data block will be updated in the background; the verification delay queue can be expanded to 8 bits or 16 bit, the extended queue can record the segment number of the data block that needs to be updated and the corresponding offset stripe number, so as to realize small-grained update.

The mirror management module is used to manage space and mirror data, including allocating space for original data, allocating space for mirror data, searching for corresponding mirror images according to original data, and elastically releasing mirror images according to user-specified priority and array capacity when space is insufficient Data space, specifically: when the user data requires less than or equal to half of the array capacity according to the verification-based RAID organization, all data has mirror redundancy; when the user data is required by the verification-based RAID organization When the space is greater than half of the array capacity, the user can specify the priority of the data, and first release the image of the data segment with lower priority; as the user data increases and reaches saturation, the array transforms into a parity-based RAID.

Described image management module has image mapping table and strip section bitmap, as shown in image mapping table 21 and strip section bitmap 22 among Fig. 2, the state of strip section bitmap record segment, a section needs 2 bit record states , there are four states in total, which are: "00" means idle; "01" means there is data; "10" means there is data and its mirror data is in a similar downward position, such as stripe segment 2 and stripe segment N +2 is adjacent downward; "11" indicates that there is data and the data has a mirror, and the mirror position needs to be looked up through the mirror mapping table; the mirror mapping table records the position of non-adjacent mirrors, and the previous item is the original data strip The segment number of the segment, the last item is the segment number of the mirrored segment.

The address conversion module is used to map the logical address into the physical address on the disk according to the segment address layout and the data layout based on the parity RAID.

The verification-based RAID module completes the data layout in the segment, the verification code calculation, and in the case of no mirror data, downgraded reading and data reconstruction.

In embodiment 2, the array construction method based on verifying RAID and adding mirror structure is applied to the array read-write system of the present invention, including four steps of address layout, data layout, data access and data reconstruction, as follows:

(1) Address layout:

(1.1) Set the number of stripes in the segment according to the number of disks M, and the number of stripes in the segment is an integer multiple of M;

(1.2) determine the segment number K according to the number of data blocks in the disk and the number of stripes in the segment, the number of segments K=the number of data blocks in the disk/segment number of stripes;

(1.3) Get N=K/2; Interleave the segments whose segment numbers are 1 to N with the segments whose segment numbers are N+1 to K: store the segment whose number is N+1 after the segment whose number is 1, and then store For the segment numbered 2, the segment numbered N+2 will be stored after the number 2, and stored interleaved in this way; if K/2 is a non-integer, it will be rounded;

(2) Data layout:

(2.1) The data in the original data segment is in the same layout as the parity-based RAID;

(2.2) On the basis of the layout of the original data segment, the mirrored data segment changes the disk number stored in the data block in the same stripe, and the disk number is uniformly shifted to the right or left; uniformly shifted to the right, that is, the disk number plus i, If the disk number after adding i is greater than M, then subtract M; uniformly shift to the left, that is, shift the disk number and subtract i, if the disk number after subtracting i is less than 0, then add M;

(3) Data reading and writing:

(3.1) Calculate the physical address of the data to be accessed; if it is a write request, also calculate the physical address of the verification information;

(3.2) Judging whether the data to be accessed has a mirror image; if there is no mirror image, then enter step (3.3), if there is a mirror image, then enter step (3.4);

(3.3) Send a read and write request to the disk pointed to by the physical address of the data; if it is a write request, update the verification information of the original data;

(3.4) Calculate the physical address of the mirrored data; if it is a write request, also calculate the physical address of the mirrored data verification data, send the write request to the disk pointed to by the corresponding data physical address and the disk pointed to by the mirrored physical address, and update The verification information of the original data and the mirrored data; if it is read, balance the load size of the disk where the original data is located and the disk where the mirrored data is located, and select the one with the smaller load to issue a read request;

(4) Data reconstruction:

If there is a faulty disk, determine whether the data on the faulty disk has a mirror image. If there is a mirror image, read the corresponding data from the mirror image and write it to the backup disk; Other data blocks calculate the data of the faulty disk and write it into the backup disk.

The verification RAID in the array construction method of embodiment 2 is RAID5, and the data layout of RAID5 is as shown in Figure 3, and Ai, Bi, Ci are data blocks, and Pi is the verification information of the ith stripe data, for example P1 is Verification information of A1, B1, and C1.

The data layout method in Embodiment 2 is shown in Figure 4: Ai, Bi, and Ci are data blocks, Pi is the verification information of the i-th stripe data, and Ai', Bi', Ci', and Pi' are respectively Ai , Bi, Ci, and Pi mirror images. The data is divided into sections. A section is divided into several stripes. The number of stripes in a section is specified by the user. In this embodiment, a section includes 4 stripes. The data layout in the original data segment is the same as the RAID5 layout, and the mirrored data segment changes the stored disk number on the basis of the original data segment layout, and uniformly shifts to the right by 1, as shown in part (a) of Figure 4; or Uniformly offset to the left by 1, as shown in part (b) of Figure 4. For example, the data originally stored on the disk numbered m is stored on the disk numbered m+1 or on the disk numbered m-1 in the mirror.

In Embodiment 2, the segment address layout method is specifically as follows: assuming that there are 2N segments, the segments numbered 1 to N are interleaved with the segments numbered N+1 to 2*N. Specifically, as shown in FIG. 5 , the segment numbered N+1 is arranged after the segment numbered 1, the segment numbered 2 is arranged after that, and the segment numbered N+2 is arranged after that, so that cross-addressing is performed. In the mirror management module, allocate space for the original data from segment number 1 to 2N to search for free allocation in sequence, allocate mirror space for the original data, and preferentially allocate the stripe segment adjacent to the physical address of the original data downward, for example, in Figure 5, the stripe segment Segment k is the original data. If a new file is written, if segment k+1 is free, segment k+1 will be allocated first. If there is data in segment k+1, view segment k +2 Is it free? Search for free space allocation sequentially. If the original data of stripe segment k is assigned a mirror image, stripe segment N+k will be allocated first. If stripe segment N+k has data, it will first search for free stripes forward. Segment allocation, if not, then look for free stripe segments later.

In Embodiment 2, the method for writing user data is specifically: for an upper-layer user write request, if the data pointed to by the write request does not have a mirror image, calculate the physical address of the original data, issue the write request and update the verification information; if the write request points to According to the data layout and mirror management, find the physical address of the original data and the physical address of the mirrored data, send a write request and update the verification information. There are two ways to update the verification information, one is background processing, when the system is idle or the load is not high, the original verification information and its mirror verification information are updated; the second is simultaneous processing, when writing the original data and the mirror image The original checksum and mirror checksum are updated at the same time when the data is updated.

More specifically, as shown in Figure 6, the user data writing specifically includes the following steps:

Step S401 calculates the physical address of the original data and the physical address of the verification information according to the data layout and the segment address layout;

Step S402 judges whether there is a mirror image, and if there is a mirror image, proceed to step S403; otherwise, proceed to step S408; said mirror image judgment is based on the numerical judgment of the stripe segment bitmap record in the mirror image management module, if it is "10" or " 11", it means there is a mirror image, if it is "01", it means there is no mirror image;

Step S403 calculates the data physical address of the image data and the physical address of the verification data thereof;

Step S404 sends a request to the disk where the original data and mirrored data are located;

Step S405 checks whether the information is updated synchronously, if yes, proceeds to step S406, otherwise, proceeds to step S407;

Step S406 updates the verification information of the original data and the image data;

Step S407 etc. When the system is idle or the disk load is not high, the verification information of the original data and the mirrored data is updated in the background; if the background is updated, the corresponding segment number and the offset stripe number in the segment will be added to the verification delay queue 11;

Step S408 sends a write request to the disk where the original data is located and updates the verification information of the original data.

The user data reading process is specifically as follows: for the upper-layer user read request, calculate the physical address of the original data; if there is a mirror image, calculate the physical address of the mirrored data, and send the read request to a disk with a smaller load; if there is no mirror image, directly send to the disk where the original data resides; in the event of a failure on the disk where the data resides, if the data is mirrored, calculate the physical address of the mirrored data, and send the user’s read request to the disk where the mirrored data resides; RAID module, downgrade read.

More specifically, as shown in Figure 7, the reading of user data specifically includes the following steps:

Step S501 calculates the original data physical address according to the data layout and segment address layout;

Step S502 judges whether there is a mirror image, if there is a mirror image, proceed to step S503, otherwise, proceed to step S507;

Step S503 calculates the data physical address of the image data;

Step S504 judges whether there is a bad disk, if there is a bad disk in the data and image disk, then proceed to step S505, otherwise, proceed to step S506;

Step S505 sends the request to the disk where the image data is located;

Step S506 balances the loads of the disks where the original data and the mirrored data reside, and selects the disk with the lighter load to send the request;

Step S507 judges whether there is a bad disk, if the original data place disk breaks down, then proceed to step S509, otherwise, proceed to step S508;

Step S508 sends the request to the disk where the original data is located;

Step S509 downgrade read, perform downgrade read according to the method based on verifying RAID, follow the method of RAID5 in this embodiment, such as in Fig. 4, if read data block C5, C disk failure, read data block A5 at this moment , B5, P5, calculate the data of C5.

In Embodiment 2, the user data reconstruction method is specifically: if a faulty disk occurs, determine whether the data on the faulty disk has a mirror image, if there is a mirror image, read the corresponding data from the mirror image and write it to the backup disk, if there is no mirror image, then based on Verify the RAID reconstruction method, read other data blocks in the stripe, calculate the data of the faulty disk, and write it to the backup disk.

More specifically, as shown in Figure 8, the reconstruction method specifically includes the following steps:

Step S601 judges whether the data has a mirror image, if there is a mirror image, proceed to step S603, otherwise, proceed to step S602;

Step S602 reads the verification information and the information of other normal disks to calculate the data on the faulty disk;

Step S603 reads the data of the faulty disk from the image;

Step S604 writes the data into the backup disk.

The array construction method of embodiment 2, by adding the mirror structure based on the verification RAID, utilizes the free space to store the mirror image of the data, improves the read and write performance of the array, alleviates the write amplification problem based on the verification RAID, and improves the performance of the array. availability and reliability.

Those skilled in the art can easily understand that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, All should be included within the protection scope of the present invention.

Claims (10)

1. a kind of array construction method that adds mirror image structure based on checking RAID, it is characterized in that, described array construction method comprises address layout, data layout, data access and data reconstruction four steps, specifically as follows: (1) Address layout: (1.1) Set the number of stripes in the segment according to the number of disks M, and the number of stripes in the segment is an integer multiple of M; (1.2) determine the segment number K according to the number of data blocks in the disk and the number of stripes in the segment, the number of segments K=the number of data blocks in the disk/segment number of stripes; (1.3) Get N=K/2; Interleave the segments whose segment numbers are 1 to N with the segments whose segment numbers are N+1 to K: store the segment whose number is N+1 after the segment whose number is 1, and then store For the segment numbered 2, the segment numbered N+2 will be stored after the number 2, and stored interleaved in this way; if K/2 is a non-integer, it will be rounded; (2) Data layout: (2.1) The data in the original data segment is in the same layout as the parity-based RAID; (2.2) On the basis of the layout of the original data segment, the mirrored data segment changes the disk number stored in the data block in the same stripe, and the disk number is uniformly shifted to the right or left; uniformly shifted to the right, that is, the disk number plus i, If the disk number after adding i is greater than M, then subtract M; uniformly shift to the left, that is, shift the disk number and subtract i, if the disk number after subtracting i is less than 0, then add M; (3) Data reading and writing: (3.1) Calculate the physical address of the data to be accessed; if it is a write request, also calculate the physical address of the verification information; (3.2) Judging whether the data to be accessed has a mirror image; if there is no mirror image, then enter step (3.3), if there is a mirror image, then enter step (3.4); (3.3) Send a read and write request to the disk pointed to by the physical address of the data; if it is a write request, update the verification information of the original data; (3.4) Calculate the physical address of the mirrored data; if it is a write request, also calculate the physical address of the mirrored data verification data, send the write request to the disk pointed to by the corresponding data physical address and the disk pointed to by the mirrored physical address, and update The verification information of the original data and the mirrored data; if it is read, balance the load size of the disk where the original data is located and the disk where the mirrored data is located, and select the one with the smaller load to issue a read request; (4) Data reconstruction: If there is a faulty disk, determine whether the data on the faulty disk has a mirror image. If there is a mirror image, read the corresponding data from the mirror image and write it to the backup disk; Other data blocks calculate the data of the faulty disk and write it into the backup disk. 2. The array construction method according to claim 1, wherein, in the step (3), when writing data, if the array space is insufficient, the mirrored data space is elastically released according to the priority specified by the user: when the user When the space required by data organization based on parity is less than or equal to half of the array capacity, all data has mirror redundancy; when user data is organized by parity-based RAID, the space required is greater than half of the array capacity , According to the priority of data specified by the user, the mirror image of the data segment with low priority is released first; as the increase of user data reaches saturation, the array transforms into a RAID based on parity. 3. the array construction method as claimed in claim 1 or 2, is characterized in that, in the described step (3), the method for judging whether there is a mirror image is specifically: according to the content judgment of the stripe segment bitmap record in the mirror image management module, "00" means free; "01" means there is data but no mirror image; "10" means there is data and its mirror data is in the downward adjacent position; "11" means there is data and the data has a mirror image, and the mirror position needs to pass Mirror map lookup. 4. The array construction method according to any one of claims 1 to 3, characterized in that there are two ways to update the verification information in the step (3): one, if the user does not require the verification data to be updated synchronously , then use the verification delay queue in the I/O module for background processing; second, synchronous processing, when writing the original data and the mirrored data, the verification information of the original data and the verification information of the mirrored data are updated synchronously. 5. The array construction method according to any one of claims 1 to 4, characterized in that, in the step (3), in the process of executing the read request, if the disk where the original data is located fails, then follow the verification RAID The method performs a read. 6. The array construction method according to any one of claims 1 to 5, characterized in that, in the step (3), during the execution of the read request, if the disk where the original data is located fails, the user read request is directly sent to Send it to the disk where the mirrored data resides. 7. A kind of array reading and writing system that adds mirror image structure based on verification RAID, it is characterized in that, described system comprises I/O module, mirror image management module, address conversion module, based on verification RAID module; The I/O module receives the upper layer read and write request, and outputs logical address information; the mirror management module judges and searches the mirror data of the original data pointed to by the address according to the logical address information, and outputs the mirror data logical address; the address conversion module Receive the logical address of the original data and mirrored data, and output the physical address of the original data and mirrored data; if it is a write request, also output the physical address of the original data and mirrored data check block; the I/O module sends The read and write requests go to the corresponding disk; the parity-based RAID module is used to execute the read request when the disk where the original data resides fails without mirroring. 8. The array read-write system according to claim 7, wherein the I/O module includes a check delay queue for updating check information of data segments in the background. 9. The array read-write system according to claim 8, wherein the verification delay queue can be extended by 8 bits or 16 bits; the extended queue is used to record data blocks that need to be updated for verification The segment number and the corresponding offset stripe number realize small-grained updates. 10. The array read-write system according to any one of claims 7 to 9, wherein the mirror image management module is used for allocating space and allocating mirror images for raw data, including a strip segment bitmap and a mirror image mapping table; by Traverse the stripe segment bitmap to find free segments and allocate them to original data and mirror data; if the space allocated for mirror data is not adjacent downward, record the non-adjacent mirror positions in the mirror mapping table.