RU2810634C2 - Method for increasing reliability of data storage systems and corresponding device - Google Patents

Abstract

FIELD: technology for computer distributed data storage systems.

SUBSTANCE: method comprises: receiving a write request by a target controller, where the write request contains data to be written and the target controller belongs to the first control device; recording, by the target controller, data to be written into a storage device of the target controller; and recording, by the target controller, data to be written into the storage devices of the mirror controllers relative to the target controller, wherein the at least one mirror controller belongs to the second control device. According to this technical solution, the data to be recorded can be recorded and stored in several control devices. When any control device fails, the host computer can still access another control device to ensure continuity of service.

EFFECT: increasing the reliability of data storage in the storage system by reducing processing, duplication and load balancing between controllers.

13 cl, 10 dwg

Description

Field of technology to which the invention relates

This application relates to the field of technology for computer distributed data storage systems, and, in particular, to a method for increasing the reliability of data storage systems and a corresponding device.

State of the art

Today, distributed data storage architecture is increasingly being applied to cloud computing platforms, platforms for processing large volumes of data, and other similar systems. Distributed storage technology means that a storage device contains multiple storage controllers so that data can be written to the device using multiple storage controllers.

When writing data, a distributed storage system typically distributes the data and writes the data to the storage devices of different controllers. In addition, multiple copies of the data written to each controller's memory are further configured to prevent unavailability or loss of data due to component failure. When the controller fails, access to the data is still possible because the specified copies are available.

Currently, distributed data storage systems use an approach in which the storage management device is separated from the storage device. If both the storage controller storing the target data and the storage controller configured with a copy of the target data fail, or the storage management device corresponding to the storage controller storing the target data fails, access to the target data will become unusable. impossible, nor will it be possible to access data in a storage device connected to said storage management device. Consequently, the work of the main information service is interrupted, and the reliability of the information storage system is reduced.

Therefore, how exactly to ensure the continuity of the main information service and improve the reliability of the storage system when the storage management device is faulty is a technical problem that urgently needs to be solved today.

Disclosure of the invention

The present application provides a method for improving the reliability of a storage system and a corresponding apparatus for ensuring the continuity of a main information service when a storage management device fails due to a malfunction, so that the method effectively improves the reliability and availability of a storage system.

According to the first aspect, a method for improving the reliability of a data storage system is proposed. The data storage system includes a first control device and a second control device. The method comprises: receiving, by a target controller, a write request, where the write request contains data to be written, wherein the target controller belongs to a first control device; recording, by the target controller, this data to be written into storage devices of one or more mirror controllers relative to the target controller, where at least one of the mirror controllers belongs to the second control device.

Optionally, after the data to be written has been written to the memory of the target controller and the memory of the mirror controller relative to the target controller, the target controller transmits a write complete message to the host computer. This write complete message is used to indicate that the data to be written has been successfully written.

In the exemplary embodiment of the present application, when the data to be written and transmitted in the write request is written, the data to be written not only needs to be written in the storage device of the target controller, but also needs to be written in the storage device of the mirror controller relative to the target controller. In addition to this, a mirror controller is a controller that is not located in the same control device as the target controller. With this approach, it is possible to ensure that the data to be written can be written to different control devices, so that when any control device fails, the host computer can still access the data to be written using another device management. This ensures continuity of the main information service and increases the reliability and availability of the data storage system.

According to the first aspect, in one possible embodiment, the first control device and the second control device are connected to a data storage device. When the first control device fails, the host computer accesses and accesses the data storage device using the second control device.

In this embodiment of the present application, the first control device and the second control device can access and access the data storage device at the same time. With this approach, it can be ensured that when any of the control devices fails, the host computer can still access and access the storage device using another control device to ensure continuity of the main information service and improve the reliability of the storage system data.

In accordance with the first aspect, in one possible implementation, the target controller and mirror controllers relative to the target controller form a redundant group of controllers. The redundant controller group contains N controllers. Here, N is an integer not less than 2. When M controllers in the redundant controller group are faulty, and the set of these M controllers contains the target controller, the primary controller selects a controller from the set of mirror controllers relative to the target controller as the new target controller, where the new target controller is configured to receiving a write request, wherein the primary controller is a controller selected from a plurality of controllers of the first control device and the second control device and configured to control the other controller; and this primary controller sets up M new mirror controllers for the new target controller, and mirrors data from the new target controller's storage device to the M new mirror controllers. Here M is an integer not less than 1 and less than N.

In this embodiment of the present application, when multiple controllers including the target controller are faulty, the primary controller may select a new target controller from a plurality of mirror controllers relative to the target controller and install new mirror controllers relative to the new target controller to ensure that the host computer can still address and access target data when multiple controllers fail. This ensures that the main information service is not interrupted and increases the reliability of the data storage system.

In accordance with the first aspect, in one possible implementation, the target controller and the mirror controllers relative to that target controller form a redundant group of controllers. This redundant controller group contains N controllers. Here, N is an integer not less than 2. When M controllers from the redundant controller group are faulty and the collection of these M controllers does not contain the target controller, the primary controller installs M new mirror controllers for the specified target controller and mirrors the data from the target controller's storage device to M new ones. mirror controllers. The primary controller is a controller selected from a plurality of controllers in the first control device and the second control device and configured to control the other controller. Here M is an integer not less than 1 and less than N.

In this embodiment of the present application, when a plurality of controllers not containing a target controller fails, the primary controller may install new mirror controllers for the target controller to ensure that the target controller can still have multiple mirror controllers, and when the multiple controllers fail , the host computer can still access and access the target data. This ensures continuity of the main information service and increases the reliability of the data storage system.

According to the first aspect, in one possible embodiment, N is 3.

According to the first aspect, in one possible embodiment, when the first control device is faulty, one mirror controller relative to the target controller located in the second control device is used as a new target controller, a mirror controller is set to the new target controller, and the mirror is reflected. data from the new target controller's storage device to the new mirror controller's storage device.

In this embodiment of the present application, the controller in each control device has a mirror controller in the other control device. When the control device in question fails, the primary controller updates and upgrades the mirror controller relative to the specified target controller located in the other control device to become a new target controller and installs mirror controllers for this new target controller. With this approach, when this control device is faulty, the host computer can still access the target data. This ensures continuity of the main information service and improves the reliability and availability of the data storage system.

In accordance with the second aspect, a data storage management system is proposed. This storage management system includes a first control device and a second control device. The first control device contains a target controller. This target controller includes a receiving module configured to receive a write request, where the write request contains data to be written; and a processor module configured to write data to be recorded into a storage device of the target controller. This processor module is further configured to write data to be written into a memory of the mirror controller relative to the target controller. The at least one mirror controller belongs to the second control device.

In accordance with the second aspect, in one possible implementation, the target controller and the mirror controllers relative to that target controller form a redundant group of controllers. This redundant controller group contains N controllers. Here N is an integer not less than 2. The storage management system contains a primary controller. This primary controller is a controller selected from a plurality of controllers in the first control device and the second control device and configured to control the other controller. Such a primary controller comprises: a fault finding module configured to: when M controllers in a redundant group of controllers are faulty and the set of these M controllers contains a target controller, select any controller from the set of mirror controllers relative to the target controller as the new target controller, where this new target controller is configured to receive the write request; and a mirroring module configured to establish M new mirror controllers for the new target controller and mirror data from the storage device of the new target controller to the M new mirror controllers. Here M is an integer not less than 1 and less than N.

In accordance with the second aspect, in one possible implementation, the target controller and mirror controllers relative to the target controller form a redundant group of controllers. This redundant controller group contains N controllers. Here N is an integer not less than 2. The storage management system further comprises a primary controller. This primary controller is a controller selected from a plurality of controllers in the first control device and in the second control device and configured to control the other controller. The primary controller includes: a fault finding module configured to: when M controllers from a redundant group of controllers are faulty and the collection of these M controllers does not contain a target controller, install M new mirror controllers for the specified target controller; and a mirror module configured to mirror data from the target controller memory to the M new mirror controllers. Here M is an integer not less than 1 and less than N.

In accordance with the second aspect, in one possible implementation, N is 3.

According to the second aspect, in one possible embodiment, when the first control device is faulty, one mirror controller that belongs to the target controller and which is located in the second control device is used as a new target controller, any mirror controller is installed for that new target controller, and mirrors the data from the new target controller's storage device to the new mirror controller's storage device.

In accordance with the third aspect, a data storage management system is proposed. This storage management system includes a first control device and a second control device. The first control device contains a target controller. This target controller contains a processor and a storage device. The processor executes computer instructions stored in the storage device to enable the target controller to implement the method of the first aspect or any one of the embodiments of the first aspect.

According to a third aspect, in one embodiment, the storage management system includes a primary controller. This primary controller is a controller selected from a plurality of controllers in the first control device and the second control device and configured to control the other controller. The primary controller contains a processor and a memory device. The processor executes computer instructions stored in the storage device to allow the primary controller to implement the method provided in accordance with the first aspect or any of the embodiments of this first aspect.

According to a fourth aspect, a computer storage medium for storing information is provided. This computer storage medium stores a computer program. When executing the computer program, the processor may implement a procedure of the method provided by the first aspect or any one of the embodiments of the first aspect.

In accordance with the fifth aspect, a computer program product is proposed. This computer program product contains computer instructions. When executing a computer program, the computer may perform a procedure of a method provided by the first aspect or any one of the embodiments of the first aspect.

Brief description of drawings

Fig. 1A is a simplified diagram of a serial connection model according to one embodiment of the present application;

Fig. 1B is a simplified diagram of a parallel connection model according to one embodiment of the present application;

Fig. 2A is a simplified block diagram of a data storage system according to an embodiment of the present application;

Fig. 2B is a simplified block diagram of a device for storing duplicate copies of data according to an embodiment of the present application;

Fig. 3 represents a simplified block diagram of an application scenario according to one embodiment of the present application;

Fig. 4 is a simplified logic diagram of a method for increasing the reliability of a data storage system according to one embodiment of the present application;

Fig. 5 represents a simplified diagram of recording multiple copies of data according to one embodiment of the present application;

Fig. 6A-1 and Fig. 6A-2 shows a simplified diagram for the case of simultaneous failure of two controllers according to one embodiment of this application;

Fig. 6B-1 and Fig. 6B-2 shows a simplified diagram for the event of a malfunction of one of the control devices according to one embodiment of this application;

Fig. 6C-1 - Fig. 6C-4 shows a simplified diagram for the case of sequential failures of controllers according to one embodiment of this application;

Fig. 7 represents a simplified diagram of a reliability model of a control device according to one embodiment of the present application;

Fig. 8 represents a simplified diagram of a reliability model of a data storage system according to one embodiment of the present application;

Fig. 9 is a simplified block diagram of a storage management system according to one embodiment of the present application; And

Fig. 10 is a simplified block diagram of another storage management system according to an embodiment of the present application.

Carrying out the invention

The following clearly describes the technical solutions of the embodiments of the present application with reference to the accompanying drawings. It will be appreciated that the embodiments described herein are only a portion and not all of the embodiments of this application. All other options obtained even by ordinary specialists in the field in question based on the options presented here without much creative effort will fall within the scope of protection of this application.

Certain terms and related technologies in this application are first described with reference to the accompanying drawings to assist those skilled in the art to better understand the present invention.

The peripheral component interconnect express (PCIe) bus is used for high-speed serial communications over a broadband, point-to-point, dual-channel connection. The attached device is allocated an exclusive channel bandwidth so that it does not share the bus bandwidth. Therefore, the data transfer rate is high. The connection between any two PCIe devices is called a link and is established from a group of one or more links.

Small computer system interface (SCSI) is an independent processor standard used for a system-level interface between a computer and a smart device (such as a hard disk drive, CD-ROM drive, printer, or scanner). The SCSI interface is a general-purpose interface. The SCSI interface bus can be connected to a host adapter and a SCSI peripheral controller. Multiple peripheral devices installed on the same SCSI bus can operate at the same time. The SCSI interface can transfer data either synchronously or asynchronously. The SCSI interface manages the data read/write operation using an independent high-speed SCSI interface card so that the central processing unit (CPU) does not have to waste time waiting. This can improve the overall functioning of the system.

Serial attached SCSI (SAS) is the next generation of SCSI interface technology that uses serial technology to achieve higher data transfer rates, improve internal space through shorter connections, and is compatible with hard disk drives with an advanced serial connection ( serial advanced technology attachment (SATA). The SAS interface has a point-to-point structure and can establish a direct connection between the drive and the controller. In addition to this, the SAS interface structure has excellent scalability and can be connected to a maximum of 16384 disk drives (disk drives). Each SAS interface has four cables - two for input and two for output. Data can be read and written at the same time.

Non-volatile memory express (NVMe) is a logical unit interface specification and uses a protocol similar to the serial ATA advanced host controller interface (AHCI) designed for flash memory devices. . The essence of the NVMe interface is to establish multiple paths between a computer and a storage device. These multiple paths are actually multiple queues. The NVMe protocol can support a maximum of 64k queues, greatly improving concurrency and performance.

Direct memory access (DMA) is an operating principle in which input/output (I/O) switching is performed in hardware. In this approach, the DMA access controller takes full control of the bus from the CPU, and direct data exchange between the storage device and the input/output (IO) device occurs without passing through the CPU. The DMA access principle can reduce CPU overhead when transferring large amounts of data. The DMA access controller is used to generate the storage access address and control the storage access procedure. The data transfer rate with this approach is high, and the CPU processor basically does not interfere and takes part only during the initialization and completion of the exchange. The CPU processor and the peripheral device operate in parallel, resulting in relatively high efficiency.

Remote direct memory access (RDMA) is a technology that bypasses the operating system kernel of a remote host computer to access data in the remote host computer's storage device. As the operating system is bypassed, the RDMA access system quickly moves data from the system to the remote computer's storage device through the communications network. Therefore, it is possible to save a large amount of CPU resources, increase throughput, eliminate the overhead of external storage duplication and context switching, and free up storage bandwidth to improve the performance and other characteristics of the proposed system.

A copy of data is a backup, redundant amount of data relative to the data being saved. More specifically, the same data is stored in different storage devices to improve system availability. If one storage device that stores the primary copy of the data fails, another storage device that stores a copy of the data can be used to retrieve the required data and thereby ensure continuity of service.

Failure rate, also called failure rate, refers to how often a technical system or its components fail. The failure rate is usually measured by the number of failures per hour and is denoted by the letter λ. Mean time between failures (MTBF) denotes the average operating time of a repairable device from the current failure to the next failure. λ = 1/MTBF. Mean time to repair (MTTR) refers to the average recovery (repair) time during which a device goes from a faulty state to a working state. Availability (A) or readiness refers to the degree to which a system, subsystem, or device is in a specified operational or deployable state at the time the task begins. In other words, availability or availability represents the proportion of time that the system is in an operational state. A = MTBF/(MTBF + MTTR). To calculate the availability or availability of a system, calculate the MTBF time for each single block (subsystem, device, etc.) of the system based on the failure rate of that block and then calculate the overall availability or availability of the system based on serial and parallel connection models. In FIG. Figure 1A shows a simplified diagram of the serial connection model. In this circuit, block 1, block 2, ..., and block N are connected in series. The availability of block 1 is A1, the availability of block 2 is A2, ..., and the availability of block N is An. Therefore, the availability (readiness) of the entire system is A = A1 * A2 * ... * An. In FIG. Figure 1B shows a simplified diagram of the parallel connection model. The parallel connection model is more complex than the serial connection model. The backup units in the parallel connection model are classified into primary and secondary backup units and load-sharing backup units. If block 1, block 2, ..., and block N in FIG. 1B are the primary and secondary backup units and the availability (availability) of each of these units is "a", the availability (availability of the entire system is A = 1 - {C * {1 - * (N + 1 - N * a)}/N + (1 - C) * (1 - ) /N}. The parameter C represents the switching frequency, namely the probability of a successful switch from the primary block to the secondary block. C=C1*C2. Here, C1 represents the failure detection rate of the primary unit, and C2 represents the failure detection rate of the secondary unit. In particular, if there is only one spare block for each block, in other words, N=1, this formula can be simplified to A = a + C * a * (1 - a). If block 1, block 2, ..., and block N in FIG. 1B are backup units with load balancing and the availability (availability) of each unit is "a", then the availability (availability) of the entire system is A = 1 - { * {1 - * (N + 1 - N * a)}/N + (1 - ) * (1 - ) / (N + 1)}. Here denotes the failure detection rate for each block.

In general, to improve system scalability, an enterprise storage system typically connects multiple storage management nodes to form a federated distributed system and provides a storage service for use. More specifically, refer to FIG. 2A, in this FIG. 2A shows a simplified block diagram of a data storage system. As shown in FIG. 2A, device 210 is a device using a storage service, such as a host or server. The computer 210 is connected to the control device 220 and the control device 230. The control device 220 includes multiple controllers. In the embodiment under consideration, four controllers are used as an example for description. These four controllers are controller 221, controller 222, controller 223 and controller 224, so these controllers are connected to each other via PCIe interfaces. The control device 230 also includes several controllers. In the embodiment under consideration, four controllers are also used as an example for description. These four controllers are controller 231, controller 232, controller 233 and controller 234, and they are also connected to each other via PCIe interfaces. To increase system reliability, these controllers are connected to each other using a dual-channel redundancy scheme. The control device 220 is also coupled to the control device 230 in a dual-lane redundant manner via a PCIe interface. The control device 220 and the control device 230 may be data storage arrays. Any controller included in the control device 220 and the control device 230 may use a central processing unit (CPU) plus memory structure. The CPU recalls a program from the storage device to carry out the corresponding function. Alternatively, the controller may be an independent control integrated circuit chip, such as a field-programmable gate array (FPGA), and the program may be written on the FPGA. The control device 220 is connected to the storage device 240 via a SAS interface. The control device 230 is connected to the storage device 250 via a SAS interface. To improve system reliability, the connection between the control device 220 and the storage device 240, and the connection between the control device 230 and the storage device 250 are implemented in a dual-channel redundant design. Storage device 240 and storage device 250 are data storage devices, such as storage arrays. The storage device 240 includes an expansion board 241, an expansion board 242, and N storage media. Here, N is a positive integer greater than 1. The expansion board 241 is responsible for managing the storage device 240. The expansion board 242 is a backup expansion board to the expansion board 241. When the expansion board 241 fails, the expansion board 242 takes over the operation of the expansion board 241 to improve the reliability of the entire system. The structure of the storage device 250 is consistent with the structure of the storage device 240 and contains an expansion board 251, an expansion board 252 and N storage media.

In the actual application, the storage system uses dual memory copy technology to ensure system reliability and controller performance and other performance. More specifically, data recorded in the storage device of each controller is backed up in the storage device of another controller belonging to the same control device. As shown in FIG. 2B, data A is stored in the controller 221, and a copy A1 of this data A is stored in the controller 222. Data B is stored in the controller 222, and a copy B1 of this data B is stored in the controller 221. Data E is stored in the controller 231, and a copy E1 of this data E is stored in controller 232. Data H is stored in controller 234, and a copy H1 of this data H is stored in controller 233. If controller 221 fails due to a malfunction, computer 210 cannot access the data stored in controller 221. More specifically, the computer 210 cannot access data A and data B1. Since the controller 222 stores a copy A1 of the data A, the computer 210 can access the data A1 by accessing and accessing the controller 222. With this approach, the service is not interrupted.

If both controller 221 and controller 222 are faulty or control device 220 is faulty, computer 210 cannot access data A and data A1, so the service may be interrupted.

In addition to this, the controller in each control device can only access the storage device connected to that control device. For example, the controller 221 in the control device 220 can only access data in the storage device 240, and cannot access the storage device 250 because the control device 220 is not connected to the storage device 250. Likewise, the controller 231 in the control device 230 can access accesses only the data in the storage device 250, and cannot access the data in the storage device 240. Therefore, when any control device fails due to power failure or hardware malfunction, the data in the controller of that control device and access to data in a storage device connected to this control device will not be possible. As a result, the service in the host computer is interrupted or completely fails. For example, if control device 220 fails, computer 210 cannot access data in device 220 and storage device 240. As a result, service to computer 210 is interrupted or fails altogether.

It should be noted that the control device and the storage device may be connected via a SAS interface line or an NVMe interface line, and each control device may be further connected to multiple storage devices. In addition, the control device and the storage device may be deployed on the same physical device. For example, storage device 240 and control device 220 may be deployed on the same physical device. Naturally, the control device and the storage device may alternatively be deployed separately on different physical devices.

To solve the above problems, this application proposes a method for increasing the reliability of a storage system and corresponding devices for ensuring the continuity of the main information service when the control device fails, thereby increasing the reliability and availability (availability) of the storage system.

In FIG. 3 shows a possible application scenario according to one of the variants of this application. In this application scenario, the host computer 310 is connected to the control device 320 and to the control device 330. The host computer 310 may be connected to the control device 320 and the control device 330 via a fiber optic transmission line, or may be connected to the control device 320 and the control device 330 via a communications network such as an Ethernet network. In addition, the control device 320 and the control device 330 have two switch boards (not shown in FIG. 3). One of these switch boards is the primary switch board, and the other switch board is the secondary switch board. All controllers in the control device 320 and in the control device 330 must be connected to the primary switch board. This primary switch board is configured to support communication between a controller in one control device and a controller in another control device, and communication between the controller in the control device and storage device 340. When the primary switch board fails, the secondary switch board takes over the operation of the primary switch board to ensuring system reliability. Control device 320 includes multiple controllers. In the present embodiment, an example is used for description in which the control device 320 includes four controllers, for example, a controller 321, a controller 322, a controller 323 and a controller 324. The control device 330 also includes several controllers. In the present embodiment, an example is used for description in which the control device 330 also includes four controllers, for example, a controller 331, a controller 332, a controller 333 and a controller 334. The connections between all controllers in the control device 320 and in the control device 330, and the connection between control device 320 and control device 330 are implemented via RDMA access channels. Data between controllers can be transferred directly to the peer end of the line without the data passing through the CPU. Of course, these connections could alternatively be implemented via PCIe interfaces. The control device 320 and the control device 330 are connected separately to the storage device 340 via RDMA access channels. In other words, the storage device 340 is shared between the control device 320 and the control device 330. More specifically, the control device 320 and the control device 330 can access and access data in the storage device 340 at the same time. The storage device 340 may be a data storage server (namely, a data storage unit) comprising a CPU 341 and a storage device 342, and include N storage media configured to store data for access by controllers. Here, N is a positive integer greater than 1. Memory 342 in memory 340 stores program instructions. The CPU 341 invokes these software instructions to control the storage device 340, such as appropriately configuring the storage device 340 or updating the storage device 340. All controllers in the control devices may be configured to receive I/O data transmitted by the host device and to determine whether based on the address passed in this data, the target controller processing this I/O data. If the target controller is a controller that receives I/O data transmitted by the host computer, the target controller directly processes the I/O data. If the target controller is not the controller that receives the I/O data sent by the host computer, the controller forwards the I/O data to the target controller, and the target controller processes the data.

It should be noted that to increase system reliability, two-channel redundancy is used for all connections. More specifically, two RDMA access channels are used for each connection between controllers, between control devices, and between control device and storage device. In addition, the host computer 310 may be further connected to multiple control devices, and the control devices may be further connected to multiple shared storage devices. In addition, each controller has at least one mirror controller, and the at least one mirror controller belongs to another control device. More specifically, data stored in each controller is backed up in another controller, and at least one controller belongs to another control device. For example, a set of mirror controllers with respect to controller 321 includes controller 322 and controller 331. If data A stored in controller 321 is backed up, controller 322 and controller 331 respectively store copy A1 and copy A2 of data A. Controller 321 and controller 322 belong to control device 320, and the controller 331 belongs to the control device 330. In this approach, data A is stored in several control devices. Mirror controllers for each controller are specified in advance. Each controller and the mirror controllers corresponding to that controller form a redundant controller group.

It is easy to understand that when a controller in a control device fails, or a control device fails, since backup data is stored in multiple control devices, the host computer can still access the backup data in another control device. This ensures continuity of service. In addition to this, several control devices share one storage device. With this approach, even if one control device fails, the host computer can still access data in the storage device using a controller from another control device to ensure that service in the host computer is not interrupted or lost altogether . This effectively increases the reliability and availability (availability) of the system.

Based on the application scenario shown in FIG. 3 will be described below with reference to FIG. 4, a method for increasing the reliability of a data storage system according to one embodiment of this application. As shown in FIG. 4, the method contains, but is not limited to, the following steps.

S410: The target controller receives the write request.

In particular, the target controller may be any controller in any control device coupled to the host computer. For example, with reference to the corresponding descriptions of FIGS. 3, the target controller may be controller 321 in control device 320. Specifically, when the target controller is a controller that currently receives I/O data transmitted by the host computer, the target controller directly receives the write request transmitted by the host computer. When the target controller is not the controller currently receiving I/O data transmitted by the host computer, the target controller accepts the write request forwarded to it by the receiving controller.

Further, this write request contains data to be written and which needs to be written by the host computer.

S420: The target controller writes data to be recorded in the storage device of the target controller.

In particular, the target controller may be a control integrated circuit chip containing a CPU and a memory. After receiving the write request, the target controller obtains, by parsing the write request, the data to be written and transmitted in the write request, and writes the data to be written in a storage device of the target controller.

S430: The target controller writes data to be written in the memory of the mirror controller relative to the target controller.

As shown in FIG. 3, in this embodiment of the present invention, each controller is provided with mirror controllers, and at least one mirror controller does not belong to the same control device to which the target controller belongs. Therefore, after writing the data to be recorded in its storage device, the target controller further writes the data to be recorded in the storage device of the mirror controller. Since these controllers are connected via an RDMA access channel, the target controller can directly write data to be written in the memory of the mirror controller relative to that target controller via the RDMA access channel.

It should be understood that data is transferred between any two controllers over an RDMA access channel, and the data does not pass through the CPU. This improves the performance of the CPU controllers and communication efficiency.

For example, in FIG. 5 shows a simplified diagram of recording multiple copies of data according to one of the variants of this application. As shown in FIG. 5, the mirror controllers for the controller 521 are the controller 522 and the controller 532. Therefore, the controller 521 first writes data A in the memory of the controller 521, and then respectively writes copies A1 and A2 of this data A in the memory of the controller 522 and in the memory of the controller 532. The mirror controllers to controller 523 are controller 524 and controller 534. After writing B data to controller 523's memory, controller 523 writes B1 copy of B data to controller 524's memory and writes B2 copy of B data to controller 534's memory. Likewise, the mirror controllers to controller 531 are controller 522 and controller 532. After writing data C to the memory of controller 531, controller 531 writes a copy C1 of that data C to the memory of controller 532, and writes a copy C2 of data C to the memory of controller 522. The mirror controllers to controller 533 are controller 524 and controller 534. After writing data D into the memory of controller 533, controller 533 writes a copy D1 of data D into the memory of controller 534, and writes a copy D2 of data D into the memory of controller 524.

In particular, in the totality of all interconnected controllers there is a primary controller. The primary controller may be any controller selected from among all the controllers, and the controller is configured to control the mirroring relationships for all controllers, in other words, which mirror controllers are associated with each controller. The other controller has a backup of this specularity ratio. When the primary controller fails and cannot operate, a new primary controller is selected from the remaining controllers to take over the work of the previous primary controller and control the mirroring ratios of all controllers. In addition to this, the set of mirroring relationships controlled by the primary controller avoids a situation where mirror controllers of too many controllers are associated with the same controller. The primary controller may set specularity ratios in accordance with a diffusion algorithm or other similar algorithm to effect load balancing among the controllers.

For example, as shown in FIG. 5, controller 521 is a target controller. Controller 522 and controller 532 are mirror controllers of this target controller. After the controller 521 writes data A to the memory of the controller 521, copies A1 and A2 of the data A are written to the memory of the controller 522 and the memory of the controller 532, respectively.

S440: The target controller transmits a write completion message to the host computer.

Specifically, after writing the data to be written to the memory of the target controller and the memories of the mirror controllers relative to the target controller, the target controller transmits a write complete message to the host computer to notify the host computer that the data to be written has been successfully written to the storage memory. device of the target controller and into the storage devices of the mirror controllers relative to this target controller.

In one possible implementation, the target controller and mirror controllers relative to the target controller form a redundant group of controllers. This redundant controller group contains N controllers. Here, N is an integer not less than 2. When M controllers in the redundant controller group are faulty, and the set of these M controllers contains the target controller, the primary controller selects one controller from the set of mirror controllers relative to the target controller as the new target controller, where this new target controller configured to receive a write request, and the primary controller is configured to receive a write request from the master device and forward the write request to the new target controller; and this primary controller installs M new mirror controllers for the new target controller and mirrors data from the new target controller's storage device to the M new mirror controllers. Here M is an integer not less than 1 and less than N.

In particular, every two controllers in all control devices periodically transmit control packets to each other to determine the corresponding statuses. If a controller does not accept a control packet sent by another controller within a specified period of time, it may determine that the other controller is faulty. In such a case, the primary controller will need to re-adjust the specularity ratios for each controller. The transmission period of control packets and the specified waiting period for receiving such a packet can be set based on current requirements. This application is not limited to this. For example, in FIG. 5, eight controllers in the control device 520 and in the control device 530 transmit control packets to one another. If controller 521 does not receive, within a predetermined period of time, a control packet sent by controller 522, controller 521 may determine that controller 522 is faulty.

As an option, the primary controller can assign priority levels to mirror controllers relative to any controller it controls. More specifically, when the target controller fails, a mirror controller with a higher priority is selected from the plurality of mirror controllers relative to the specified target controller as a new target controller based on this priority. As an option, the primary controller can install a mirror controller that is part of the same control device as the target controller as a higher priority controller. Of course, the higher priority controller may alternatively be specified based on another condition, such as the physical status and backup power supply status of the corresponding controller. This is not limited in this application. For example, in FIG. 5, both controller 522 and controller 532 are mirror controllers to controller 521, but the priority of controller 522 is higher than the priority of controller 532. Therefore, when controller 521 fails, the primary controller will select controller 522 as the new target controller. In an alternative embodiment, the primary controller randomly selects a controller from the plurality of mirror controllers relative to the target controller as the new target controller. The method used by the primary controller to determine a new target controller is not limited to this application.

Next, when M controllers including the target controller are faulty, the primary controller determines a new target controller from the set of remaining mirror controllers, and installs M new mirror controllers for the new target controller so that the number of controllers in the redundant controller group remains the same as before ( namely, N controllers), and mirrors the data from the new target controller to M new mirror controllers. In addition, the primary controller may receive a write request from the host computer and then forward the write request to a new target controller. The new target controller writes the write-to-be-written data from the write request to the storage device of this new target controller and to the storage devices of the mirror controllers (including the M new mirror controllers) relative to the new target controller.

Optionally, the value of N is 3. More specifically, the target controller has two mirror controllers, and wherein the target controller and at least one of the mirror controllers are deployed in different control devices.

In one possible implementation, the target controller and mirror controllers relative to this target controller form a redundant group of controllers. This redundant controller group contains N controllers. Here, N is an integer not less than 2. When M controllers in a redundant controller group are faulty, and the collection of these M controllers does not contain the target controller, the primary controller installs M new mirror controllers for the specified target controller and mirrors the data from the target controller's storage device to M new mirror controllers. Here M is an integer not less than 1 and less than N.

In particular, when a set of M controllers from the redundant controller group that does not contain the target controller fails, the primary controller updates the global mirroring relationships. More specifically, the primary controller directly installs M new mirror controllers for the specified target controller to ensure that the number of controllers in the redundant controller group is still N, and mirrors data from the target controller to M new mirror controllers.

Optionally, the value of N is 3. More specifically, the target controller has two mirror controllers such that the target controller and at least one of the mirror controllers are deployed in different control devices.

For example, refer to FIG. 6A-1 and Fig. 6A-2. As shown in FIG. 6A-1 and Fig. 6A-2, when the control device 520 and the control device 530 are operating normally, the controller 521, the controller 522, and the controller 532 form a redundant controller group and store data A and copies A1 and A2 of the data A, respectively. The controller 522, the controller 531, and the controller 532 form another redundant group of controllers and store respectively data C and copies C1 and C2 of this data C. For data A, the target controller of this data A is controller 521. Because controller 522 and controller 532 store copies A1 and A2 of data A when controller 521 and controller 522 fail at the same time, the primary controller selects controller 532 as the new target controller. In other words, the data stored in the controller 532 is changed and turned into data A. The primary controller resets the new specularity ratio. After controller 532 is defined as the new target controller, copies A1 and A2 of data A are mirrored to controller 531 and controller 523. In other words, controller 523 and controller 531 are mirror controllers to controller 532. Likewise, for data C, the target controller of this data C is controller 531. Because controller 522 maintains a copy C2 of data C, when this controller 522 fails, the primary controller sets up a new mirror controller for target controller 531 to establish a new mirroring relationship. The primary controller mirrors a copy of data C to controller 524. In other words, controller 524 and controller 532 are mirror controllers to target controller 531.

It should be understood that the above description uses an example in which each target controller has only two mirror controllers. Of course, each target controller can have more mirror controllers. The specific implementation of logic and processes in the target controller is the same as described above. For the sake of brevity, the details will not be described here again.

It can be understood that when two controllers fail at the same time, since the target data (eg, data A) is backed up and copies of the target data are stored in three different controllers, the host computer can still access the target data. This ensures the continuity of the main information service and increases the reliability and availability of the entire data storage system. In addition, the primary controller in the control device may further re-establish the new mirroring relationship and define the controller that has not failed as the new target controller or mirror controller to ensure that the target data and copies of that target data can be stored in three different controllers.

In one embodiment, when the first control device is faulty, one mirror controller related to the target controller and located in the second control device is used as a new target controller, one of the mirror controllers is set to this new target controller, and the data from the storage device the new target controller is mirrored into the memory of the new mirror controller.

In particular, when the first control device is faulty, the primary controller needs to update the global specularity ratio again. Each controller in the first control device has at least one mirror controller in the second control device. Therefore, the primary controller may select one of the mirror controllers in the second control device as a new target controller, then install a new mirror controller for the new target controller, and mirror data from the storage device of the new target controller to the storage device of the new mirror controller. In the exemplary embodiment of the present invention, the controller in the first control device or the controller in the second control device may be selected as the primary controller. A method for selecting a primary controller in a distributed system is proposed in the prior art and the details are not described herein.

For example, refer to FIG. 6B-1 and Fig. 6B-2. As shown in FIG. 6B-1 and Fig. 6B-2, when the control device 520 and control device 530 operate normally, for data A, the target controller for that data A is controller 521, and controller 532 stores a copy A2 of that data A, so that controller 532 is a mirror controller to controller 521. For data B, the target controller for that data B is controller 523, and controller 534 maintains a B2 copy of that data B, so that controller 534 is a mirror controller to controller 523. When control device 520 fails, for data A, the primary controller uses controller 532 as the new target controller, the data stored in controller 532 is changed to data A, and the primary controller re-establishes a new mirroring ratio and mirrors copies A1 and A2 of data A to controller 531 and controller 533, respectively. For data B, the primary controller uses controller 534 as the new target controller, resets the new mirroring ratio, and mirrors copies B1 and B2 of data B to controller 531 and controller 533, respectively. For data C, the primary controller directly sets a new mirror ratio, and mirrors a copy C2 of data C to controller 534. For data D, the primary controller also directly sets a new mirror ratio, and mirrors a copy D2 of data D to controller 532.

It should be understood that the above description only uses two control devices as an example. If there are more than two control devices, the specific implementation of the logic is consistent with the material presented above, so the details will not be described here again.

It can be understood that when any control device fails, since each of the controllers in that device has at least one mirror controller in the other control device, in other words, all the data from the failed control device in question is reserved in the other control device, the host computer can still access data from a failed control device by accessing and accessing another control device. This ensures continuity of the main information service and increases the reliability and availability of the data storage system.

In a scenario in which the first control device contains four controllers and the second control device contains four controllers, the method proposed in the present invention can maintain service continuity when seven controllers fail. For details, see the embodiment shown in Fig. 6C-1 - Fig. 6C-4. As shown in FIG. 6C-1 - Fig. 6C-4, when the control device 520 and control device 530 are operating normally, the controller 521 stores data A, and the controller 522 and controller 532 respectively store copies A1 and A2 of data A. The controller 523 stores data B, and the controller 524 and controller 534 respectively store copies B1 and B2 of data B. Controller 531 stores data C, and controller 532 and controller 522, respectively, store copies C1 and C2 of data C. Controller 533 stores data D, and controller 534 and controller 524, respectively, store copies D1 and D2 of data D. When controller 521 fails, the primary controller sets a new mirroring relationship, uses controller 522 as the target controller for data A, and mirrors A1's copy of data A to controller 523. The mirroring relationship between the other controllers remains unchanged. Then the 522 controller also fails. The primary controller uses controller 523 as the target controller for data A, and mirrors copy A1 of data A in controller 524. Copy C2 of data C is mirrored in controller 524. The mirroring relationship between the other controllers remains unchanged. Further, the 523 controller also fails. This primary controller uses controller 524 as the target controller for data A, and mirrors copy A1 of data A in controller 534. The primary controller uses controller 524 as the target controller for data B, and mirrors copy B1 of data B in controller 532. Relationship mirroring between other controllers remains unchanged. Further, the 524 controller also fails. The primary controller uses controller 534 as the target controller for data A, and mirrors a copy A1 of data A in controller 531. The primary controller uses controller 532 as the target controller for data B, and mirrors a copy B1 of data B in controller 531. The primary controller re-mirrors a copy C2 of data C in controller 534. The primary controller mirrors a copy D2 of data D in controller 532. The other mirroring relationship remains unchanged. Further, controller 531 also fails. The primary controller re-mirrors A1's copy of data A into controller 533. The primary controller re-mirrors B1's copy of B's data into controller 533. The primary controller uses controller 532 as the target controller for C's data, and mirrors C1's copy of C's data into controller 533. The other specularity ratio remains unchanged. Further, the 532 controller also fails. In this case, only two controllers remain, namely, controller 533 and controller 534. Therefore, according to the mirroring relationship controlled by the primary controller, each target controller has only one mirror controller. The primary controller uses controller 533 as the target controller for data B. Controller 533 is also used as the target controller for data C and the target controller for data D. Since controller 534 stores a copy B1 of data B, a copy C1 of data C, and a copy D1 of data D, the primary controller does not need to set a new specularity ratio. Further, the 533 controller also fails. Of the entire set of controllers, only controller 534 remains. All data can only be stored in controller 534, and backup copies cannot be created. In other words, the primary controller cannot install a new mirror controller, and the 534 controller is used as the target controller for data A, data B, data C, and data D.

It can be understood that when the controllers fail one after another, the updated mirroring ratio has to be continuously established through continuous mirroring to ensure that the host computer can normally access the target data. Even if only one controller is running, the host computer can still access the target data. This ensures continuity of the main information service.

It should be noted that the above discussion simply uses an example in which each controller fails sequentially to describe changes in the specularity ratio between controllers. Of course, there is also an application scenario in which multiple controllers fail at the same time, or an application scenario in which multiple controllers fail simultaneously or sequentially. The specific logic for changing the specularity ratio of the controllers is consistent with the above. For the sake of brevity, the details will not be described here again.

In addition, to better describe the impact of this application on storage system availability, an example storage system containing two management devices and eight controllers is used. In FIG. Figure 7 shows a simplified diagram of the control device reliability model. As shown in FIG. 7, the model contains four parallel controllers, two parallel redundant power supplies, two parallel power modules, two parallel fans, one backplane and two parallel control modules, which are connected in series. The availability (readiness) of each control device in the scenarios shown in Fig. 2B and FIG. 5 are calculated separately. It should be noted that in the above two scenarios, the other components are consistent and the connection ratios are also consistent. Therefore, the availability (readiness) of the control device can only be calculated based on the corresponding redundancy of the controllers. Assuming that the availability (availability) of each controller is 0.999989496, the availability of each control device shown in FIG. 2B is calculated using Formula 1. This Formula 1 is as follows:

A = (a + C * a * (1 - a)) * (a + C * a * (1 - a)) formula 1

In Formula 1, “a” denotes the availability of each controller. Parameter C indicates the switching frequency. The value of availability (readiness) obtained as a result of calculations is equal to 0.999998950.

The availability of each control device shown in FIG. 5 is calculated using formula 2. This formula 2 is as follows:

A = 1 - {C * {1 - * (N + 1 - N * a)}/N + (1 - C) * (1 - ) /N} formula 2

In Formula 2, N denotes the number of redundant controllers (namely, the number of mirror controllers) for each controller. The availability (readiness) of the control device, obtained as a result of calculations, is equal to 0.999999475.

It can be understood that the availability of the control device shown in FIG. 5 is higher than the availability of the control device shown in FIG. 2B. Therefore, the reliability of the control device shown in FIG. 5, above.

Next, in FIG. Figure 8 shows a simplified diagram of the reliability model of a data storage system. As shown in FIG. 8, the model contains two parallel control devices, X serial hard drives and Y serial disk groups. These two management devices, X hard drives and Y disk groups are connected in series. Availability of the entire storage system shown in FIG. 2B, and the system shown in FIG. 5 is calculated using the model shown in FIG. 8. It should be understood that in the scenarios shown in FIGS. 2B and FIG. 5, other matched components can be used and the connection ratios are also matched. Therefore, the availability (availability) of the entire storage system can be calculated based only on the management devices. It should be noted that in the scenario shown in FIG. 2B, each control device is independent of other control devices. More specifically, when one of the control devices fails, the entire storage system cannot operate. Therefore, it is assumed that a serial connection model is applicable to the control devices, and the availability of the storage system shown in FIG. 2B is calculated using Formula 3. This Formula 3 is as follows:

A = A1 * A2 formula 3

In Formula 3, the values A1 and A2 indicate the availability of each control device. Since all control devices are the same, the values of A1 and A2 are the same and equal to 0.999998950. Therefore, the availability of the entire storage system is 0.999997900.

However, in the scenario shown in FIG. 5, each control device is associated with another such device. Even if one of these control devices fails, the other control device can still maintain access to the host computer so that the entire storage system can operate normally. Therefore, it is assumed that the control devices comply with the parallel connection model, and the availability of the storage system shown in FIG. 5 is calculated using formula 4. Formula 4 is as follows:

A = a + C * a * (1 - a) formula 4

In Formula 4, “a” denotes the availability of each control device. Parameter C indicates the switching frequency. The availability of the entire data storage system obtained as a result of calculations is 0.999999475.

It can be understood that the availability of the storage system shown in FIG. 5 is much higher than the availability of the storage system shown in FIG. 2B. In other words, the solutions provided by embodiments of the present application can effectively increase the availability of control devices and storage systems, improve the reliability of the storage system, and ensure the continuity of the main information service.

Methods according to embodiments of the present application are described in detail above. To facilitate the implementation of solutions according to embodiments of this application, appropriate apparatus related thereto is used to participate in the implementation of these solutions, as will be shown below.

In FIG. 9 shows a simplified block diagram of a data storage management system according to one of the variants of this application. As shown in FIG. 9, the storage management system 900 includes a first control device 9100 and a second control device 9200. The first control device 9100 includes a target controller 9110. This target controller 9110 includes a receiving module 9111 and a processing module 9112.

The 9111 receiving module is configured to receive a write request. A write request contains data to be written.

The processor module 9112 is configured to write data to be written to a memory device of the target controller 9110.

The processor module 9112 is further configured to write data to be written to a memory of the mirror controller relative to the target controller 9110. The at least one mirror controller belongs to the second control device 9200.

In one possible implementation, the target controller 9110 and the mirror controller relative to the target controller 9110 form a redundant group of controllers. This redundant controller group contains N controllers. Here, N is an integer not less than 2. The storage management system 900 includes a primary controller 9210. This primary controller 9210 is a controller selected from a plurality of controllers in the first control device 9100 and the second control device 9200 and configured to control the other controller. The primary controller 9210 includes: a fault finding module 9211, configured to: when M controllers in a redundant group of controllers fail and the set of these M controllers contains a target controller, select a controller from the set of mirror controllers relative to the target controller as the new target controller, where this new target controller is configured to receive a write request; and mirroring module 9212 is configured to establish M new mirror controllers for the new target controller and mirror data from the new target controller's storage device to the M new mirror controllers, where M is an integer not less than 1 and less than N.

In one possible implementation, N is 3.

In one possible implementation, when the first control device 9100 fails, one of the mirror controllers corresponding to the target controller 9110 and located in the second control device 9200 is used as a new target controller, installing any mirror controller for the new target controller and mirroring the data from the storage device of the new target controller into the storage device of the new mirror controller.

It should be understood that the target controller and primary controller structures described above, as well as the process for improving the reliability of the storage management system, are only examples and do not constitute specific limitations. Modules can be added to the structure of the target controller and the primary controller as needed, modules can be deleted, or any combination of modules can be made. In addition, operations and/or functions of modules in the target controller and the primary controller are respectively used to carry out the corresponding procedures of the method described in FIG. 4. For the sake of brevity, the details will not be described here again.

In FIG. 10 is a simplified block diagram of another storage management system according to one embodiment of the present application. As shown in FIG. 10, the storage management system 100 includes a first control device 1100 and a second control device 1200. The first control device 1100 includes a target controller 1110. This target controller 1110 includes a processor 1111, a communication interface 1112, and a storage device 1113. The processor 1111, a communication interface 1112, and storage device 1113 is connected via an internal bus 1114.

In one exemplary embodiment, storage management system 100 includes a primary controller 1210. This primary controller 1210 is a controller selected from a plurality of controllers in the first control device 1100 and the second control device 1200 and configured to control the other controller. The primary controller 1210 includes a processor 1211, a communications interface 1212, and a storage device 1213. The processor, communications interface 1212, and storage device 1213 are connected via an internal bus 1214.

It should be understood that the function performed by the receiving module 9111 may be particularly implemented by the communication interface 1112, the function performed by the processor module 9112 may be particularly implemented by the processor 1111, and the function performed by the troubleshooting module 9211 and mirroring module 9212 may be specifically implemented by processor 1211.

Memory 1113 may be configured to store program code and data such that processor 1111 recalls program code and data stored in memory 1113 to perform the functions of receiver module 9111 and/or processor module 9112. Processor 1111 may recall program code and data. , stored in the storage device 1113, to carry out the corresponding procedure of the method shown in FIG. 4.

Memory 1213 may be configured to store program code and data such that processor 1211 calls up program code and data stored in memory 1213 to perform the functions of troubleshooting module 9211 and/or mirroring module 9212. Processor 1211 may call program code and data stored in memory 1213 to perform the corresponding procedure of the method shown in FIG. 4.

It should be noted that each controller, target controller 1110 and primary controller 1210, may be implemented using a logic device with embedded processor logic, such as an FPGA or digital signal processor (DSP). In addition, target controller 1110 and primary controller 1210 may contain more or fewer components than shown in FIG. 10, or may have different component configurations.

It should be understood that data processing procedures for improving the reliability of a storage management system comprising a target controller 1110 and a primary controller 1210 shown in FIG. 10 can be found in the corresponding descriptions of FIGS. 1 - Fig. 8. For the sake of brevity, the details will not be described here again.

One embodiment of the present application further provides a computer-readable medium for storing data. This computer-readable storage medium stores instructions. When a computer or processor executes these instructions, it may perform some or all of the steps of any one of the methods described in the above embodiments.

One embodiment of the present application further proposes a computer program. This computer program contains commands. When running this program, the computer may perform some or all of the steps in any one of the methods to improve the reliability of the storage system.

In the above options, the description of each option focuses on the relevant point. For parts not described in detail in any embodiment, refer to the corresponding descriptions in other embodiments.

It should be noted that for the sake of brevity of description, the variants of methods discussed above are presented in the form of a sequence of actions. However, one skilled in the art will understand that this application is not limited to the procedures described herein, since some steps may be performed in a different order or simultaneously according to this application. In addition, one skilled in the art will understand that the embodiments shown herein are all exemplary embodiments and contain actions and modules not necessarily required by the present application.

In some embodiments proposed herein, it should be understood that the devices described herein may be implemented in other ways. For example, the device variants described are just examples. For example, blocking is simply a division of logical functions and may be different in the actual implementation. For example, some blocks or components may be combined or integrated into another system, or some features may be ignored or not implemented. In addition, the relationships shown or discussed, direct connections, or data connections may be implemented through some interfaces. Indirect connections or data connections between devices or units may be implemented electronically or in other forms.

The blocks described above as individual parts may or may not be physically separate, and the parts represented as blocks may or may not be physical units, may be located in one location, or may be distributed over multiple units in a communications network. Some or all of the blocks may be selected based on actual requirements to achieve the objectives of the technical solutions of the options.

In addition, the functional blocks in embodiments of the present application may be integrated within a single processing unit, or each of such blocks may exist physically separately, or two or more blocks may be integrated within a single unit. The integrated block can be implemented in hardware form or in the form of a software function block.

Claims (29)

1. A method for storing data in a data storage system comprising a first control device (220) and a second control device (230), wherein a primary controller is selected from controllers of the first control device (220) and a second control device (230) and is configured to control the others controllers, and the mirroring ratio of all controllers indicates which controllers are mirrored for each controller, containing stages in which: receiving, with the help of the target controller (221), a write request, wherein said write request contains data to be written, and the target controller belongs to the first control device; writing, with the help of the target controller, data to be recorded into a storage device of the target controller; And writing, with the help of the target controller, data to be written into the storage devices of the mirror controllers (231) relative to the target controller, wherein at least one mirror controller belongs to the second control device; and when the first control device (220) and the second control device (230) are connected to the data storage device (340), the method comprises the step of: obtaining, when the first control device is faulty, accessing the data storage device using the second control device and updating, using the primary controller, the mirroring ratios for all controllers. 2. The method according to claim 1, further comprising, when the target controller (221) and the mirror controllers (231) relative to the target controller form a redundant group of controllers, and the redundant group of controllers contains N controllers, where N is an integer not less than 2, step, on which: selecting, by the primary controller, when M controllers from the redundant group of controllers fail and the plurality of said M controllers contains a target controller, one of the controllers from the plurality of mirror controllers relative to the target controller as a new target controller, wherein said new target controller is configured to receive record request; And establish, with the help of the primary controller, M new mirror controllers for the new target controller, and mirror data from the storage device of the new target controller to M new mirror controllers, where M is an integer greater than or equal to 1 and less than N. 3. The method of claim 1, further comprising: when the target controller (221) and the mirror controllers (231) with respect to the target controller form a redundant controller group, said redundant controller group contains N controllers, where N is an integer greater than or equal to 2, step , on which: set, using the primary controller, when M controllers from the redundant group of controllers have failed and the set of said M controllers does not contain the target controller, M new mirror controllers for the target controller, and mirroring data from the target controller's storage device to M new mirror controllers, where M is an integer greater than or equal to 1 and less than N. 4. Method according to claim 2 or 3, in which N is 3. 5. The method according to claim 1, in which when the first control device (220) fails, one of the mirror controllers (231) related to the target controller (221) and located in the second control device (230) is used as a new target controller, installing a mirror controller for the new target controller, and mirroring data from the storage device of the new target controller to the storage device of the new mirror controller. 6. A storage management system (210) comprising a first control device (220) and a second control device (230), wherein a primary controller is selected from the controllers of the first control device (220) and the second control device (230) and is configured to control the others controllers, and the mirroring ratio of all controllers indicates which controllers are mirrored for each controller, with the first control device containing a target controller containing: a receiving module (9111) for receiving a write request, wherein the write request contains data to be recorded; And a processor module (9112) for writing data to be recorded into a memory of the target controller, wherein the processor module is further configured to write data to be recorded into a memory of the mirror controller (231) relative to the target controller, and at least one mirror controller belongs to the second control device; wherein the first (220) control device and the second control device (230) are connected to the data storage device (340), and the system is configured to: obtaining, when the first control device is faulty, accessing the storage device using the second control device and updating, using the primary controller, the mirroring ratio for all controllers. 7. The data storage management system (210) according to claim 6, in which the target controller (221) and the mirror controller (231) with respect to the specified target controller form a redundant group of controllers containing N controllers, where N is an integer greater than or equal to 2, wherein the primary controller contains: a fault finding module configured to select, when M controllers from a redundant group of controllers fail and the plurality of said M controllers contains a target controller, one of the mirror controllers relative to said target controller as a new target controller, wherein said new target controller is configured to receive record request; And a mirroring module configured to install M new mirror controllers for a new target controller, and mirror data from the storage device of the new target controller to M new mirror controllers, where M is an integer greater than or equal to 1 and less than N. 8. The storage management system (210) according to claim 6, wherein the target controller (221) and the mirror controllers (231) relative to the specified target controller form a redundant group of controllers containing N controllers, where N is an integer greater than or equal to 2, wherein the primary controller contains: a fault finding module for installing, when M controllers from a redundant group of controllers fail and the set of said M controllers does not contain a target controller, M new mirror controllers for the target controller; And a mirroring module for mirroring data from the target controller's storage device to M new mirror controllers, where M is an integer greater than or equal to 1 and less than N. 9. Data storage management system (210) according to claim 7 or 8, wherein N is 3. 10. The storage management system (210) of claim 6, wherein when the first management device (220) fails, one of the mirror controllers (231) related to the target controller (221) located in the second device (230 ) controls are used as a new target controller, set the mirror controller as a new target controller, and mirror data from the storage device of the new target controller to the storage device of the new mirror controller. 11. A data storage management system (210) comprising a first control device (220) and a second control device (230), wherein the first control device comprises a target controller (221) containing a processor and a storage device, wherein the processor is configured to execute computer commands stored in the storage device to cause execution, by the target controller, of the method according to any one of claims. 15. 12. The data storage management system (210) of claim 11, further comprising a primary controller (221), which is a controller selected from a plurality of controllers of the first control device (220) and the second control device (230), and configured to control the other controller (231), wherein the primary controller includes a processor and a storage device, and the processor is configured to execute computer instructions stored in the storage device to enable the primary controller to perform the method according to any one of claims. 15. 13. A computer storage medium storing a computer program that causes, when executed by a processor, the implementation of the method according to any one of claims. 15.