High-speed centralized storage active-active method and equipment
By using cache modules and management modules in a multi-control architecture, the problem that parity disk arrays cannot be written simultaneously in a multi-control environment is solved, the high performance and reliability of the storage system are achieved, and the performance and reliability of the storage system are improved. sex.
Patent Information
- Application Number
- CN202411137523.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-19
AI Technical Summary
In dual-live scenarios, parity disk arrays cannot support multiple controllers to write at the same time, resulting in data verification calculation conflicts and cache synchronization problems, affecting storage system performance.
Using cache module as a high-performance cache space and management module. In the multi-control architecture, the mirrored disk array is used as the cache module of the parity disk array. The address space is managed through the management module, avoiding bandwidth bottlenecks, and achieving fullness of multi-control hardware resources. Utilize to ensure synchronization and verification consistency of data writing.
Implement multiple controllers to read and write data at the same time, avoiding data link delay and bandwidth bottlenecks, improving storage system performance, solving the write amplification problem of parity disk arrays in multi-control environments, and improving the reliability of storage systems. performance and performance.
Smart Images

Figure CN119088624A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of computer storage, and in particular relates to a high-speed centralized storage active-active method and device. Background Art
[0002] Due to the high reliability requirements of enterprises for important data storage, dedicated storage devices usually use RAID disk array solutions to prevent damage to a single disk. Common types are mirroring (RAID1) and parity (RAID4, RAID5, RAID6, etc.). Figure 1 , using two disks (such as Disk 1 and Disk 2) to store the same content to ensure that the data remains intact when any one disk is damaged. The disadvantage is that half of the storage space is wasted, which brings greater cost pressure. Figure 2 , by using a certain amount of space redundancy to ensure data reliability, the redundant data P needs to be obtained through XOR operation with the data of other disks. When any of the disks is abnormal, the damaged disk data can also be restored through XOR operation with other disks, such as Disk 1...Disk n, Disk n+1. Since the parity check scheme requires less additional storage space, it is widely used, but the disadvantage is that it will cause write amplification problems. For example, when writing Block 1, it is necessary to read out other data blocks of the same strip of raid at the same time to recalculate the check data P, and then write Block 1 and the check data P again. Figure 1 , 2 In the table, Disk n indicates the nth disk, Block n indicates the nth disk sector, and n is a positive integer.
[0003] In addition, to ensure business continuity, dedicated storage devices usually use redundancy mechanisms to eliminate single points of failure and ensure high availability of storage services, such as dual-controller architecture (referred to as dual-controller architecture). Depending on demand, it can be further upgraded to a multi-controller architecture such as four-controller or eight-controller architecture. Figure 3 As shown in the figure, the disk (and the disk array using the disk group) can be visible to multiple controllers (primary controller and backup controller, i.e., controller A and controller B) at the same time. Each controller reads and writes data downward and provides storage services upward. The host uses multipath technology to aggregate the same volume from different controllers.
[0004] In the active-active mode, theoretically, users can read and write the same storage resource through all controllers at the same time, making full use of the computing resources of all controllers. However, the read and write process in the active-active scenario is relatively complicated due to the following issues. Specifically:
[0005] First, the industry usually uses memory in the controller as a cache to speed up read and write speeds and reduce the write amplification problem of parity disk arrays. However, when one controller rewrites the cache, the other controllers cannot perceive it, so there are cache synchronization problems, such as Figure 4 shown.
[0006] Second, the parity disk array cannot support simultaneous writing by multiple controllers, because the writing of any data block of the parity disk array requires other data blocks of the same stripe to participate in the calculation of the verification data block. If multiple controllers write different data blocks of the same stripe at the same time, there will be a data verification calculation conflict problem. Summary of the invention
[0007] The purpose of the present invention is to address the above-mentioned problems and propose a high-speed centralized storage active-active method and device, which solves the problem that the parity disk array cannot be written simultaneously in a multi-control environment by using a cache module as a high-performance cache space and management module, thereby avoiding bandwidth bottlenecks, making full use of multi-controller hardware resources, and maximizing the performance of the storage system.
[0008] To achieve the above object, the technical solution adopted by the present invention is: A high-speed centralized storage active-active method proposed in the present invention is applied to a multi-controller architecture, where the multi-controller architecture includes multiple controllers. The high-speed centralized storage active-active method includes the following steps: S1. Using a mirror disk array as a cache module of a parity disk array, the cache module is used to temporarily store data when at least two controllers of a multi-controller architecture simultaneously write data to the parity disk array, the parity disk array is used to store data, and both the mirror disk array and the parity disk array include a plurality of disks; S2. One of the controllers in the multi-controller architecture is used as the main controller, and the other controllers are used as standby controllers. Data read / write modules and management modules are deployed on each controller. Only the main controller runs the data read / write modules and management modules at the same time, and the standby controller only runs the data read / write module. The data reading and writing module is used to receive the data reading and writing request from the host and execute the data reading and writing task accordingly; The management module is used for managing the cache module and the address space. The address space is used to represent the correspondence between the logical address and the physical address of the data. S3, using the data reading and writing module and the management module to execute the data reading and writing tasks according to the data reading and writing requests sent by the host, as follows: Data writing: Each data read-write module receives a data write request sent from the host end, and notifies the management module that there is data to be written to the disk, and requests the physical address space from the management module. The management module allocates the physical address space according to the logical address and length of the written data, and returns the physical address corresponding to the physical address space to the corresponding data read-write module. Each data read-write module writes the data to the corresponding physical address space according to the returned physical address. After the writing is completed, the management module is notified to update the metadata and save the updated metadata in the mirror disk array and the parity disk array. After the metadata is updated, the management module notifies the corresponding data read-write module. After receiving the notification, the data read-write module returns to the host end that the data write request has been completed, and returns the physical address corresponding to the physical address space to the corresponding data read-write module. Specifically: when there is simultaneous writing by the controller, the management module returns that the physical address of the corresponding data read-write module belongs to the mirror disk array, otherwise, the management module returns that the physical address of the corresponding data read-write module belongs to the parity disk array; Data read: Each data read / write module receives a data read request from the host, and provides the logical address and length of the data to the management module. The logical address of the data is used as the input of the management module, and the physical address of the data is obtained from the management module. The data is read from the corresponding mirror disk array or parity disk array according to the physical address. After the read is completed, the management module is notified to update the metadata and return to the host. S4. When the main controller crashes or is abnormal, one of the backup controllers starts the management module to take over the work of the main controller; when the backup controller crashes or is abnormal, the main controller works normally; when there is an abnormality in the link between the controller and the host, no operation is required, and other controllers work normally.
[0009] Preferably, the cache module is also used to temporarily store the data when the length of the written data is less than a preset threshold, and when the length of the written data is less than the preset threshold, the management module returns the physical address of the corresponding data read and write module belonging to the mirror disk array, otherwise, the management module returns the physical address of the corresponding data read and write module belonging to the parity disk array.
[0010] Preferably, the management module is also used to move data from the cache module to the parity disk array when the cache space of the cache module is insufficient, and update the metadata after the migration is completed, and save the updated metadata in the mirror disk array and the parity disk array.
[0011] Preferably, one of the standby controllers starts the management module to take over the work of the main controller, specifically: The standby controller loads and restores the metadata stored in the mirror disk array and the parity disk array to the started management module and then restarts the operation.
[0012] Preferably, the management of the cache module adopts the bcache cache model or the ocf cache model; the management of the address space adopts the FTL model.
[0013] Preferably, the disk is a dual-port disk or a disk cabinet that can be shared by multiple controllers.
[0014] Preferably, the multi-controller architecture is a dual-controller architecture, a quad-controller architecture, or an eight-controller architecture.
[0015] A high-speed centralized storage active-active device includes a host, a multi-controller architecture, a mirror disk array and a parity disk array. The multi-controller architecture includes multiple controllers. The mirror disk array and the parity disk array are connected to each controller. Each controller is also connected to the host. Metadata interaction is performed between the controllers through a low-latency link channel. The high-speed centralized storage active-active device is used to execute any of the high-speed centralized storage active-active methods mentioned above.
[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention optimizes the hardware architecture and cooperates with software to realize the active-active storage function. Multiple controllers can read and write data and complete data requests at the same time, and the data does not need to be forwarded. Only a small amount of metadata interaction is required. Therefore, there is no need for a high-speed link between controllers for forwarding communication. Only a low-latency link is required. Bandwidth bottlenecks and the problem of longer data links and larger data delays caused by forwarding data can be avoided. Compared with existing multi-controller storage devices, the upstream and downstream link bandwidths of each controller can be utilized, and the performance can be doubled, thereby fully utilizing the hardware resources of the multi-controllers and maximizing the performance of the storage system. Specifically: by using the cache module as a high-performance cache space and management module, the problem that the parity disk array cannot be written simultaneously in a multi-control environment is solved, and the active-active storage is better realized. The address space can be managed through the management module to achieve isolation of the logical address and the physical address of the storage, which is helpful to improve confidentiality and facilitate global wear leveling and global anti-wear leveling. For example, when selecting a physical address, you can give priority to disks that are used less frequently. Normal wear leveling is a function within a single disk and cannot achieve wear leveling between disks. In addition, when managing the address space, the management module will cache small data write requests in the high-speed mirror disk array first. When the cached data accumulates to a certain amount, the data will be merged into a large data and written to the parity disk array. This can solve the write amplification problem caused by writing small data in the parity disk array, and can also make full use of the high-speed bandwidth of the mirror disk array, so that the number of read and write (I / O) operations per second for writing small data can be increased by several times to hundreds of times, which is a significant improvement. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1It is a structural schematic diagram of a mirror disk array in the prior art; Figure 2 It is a structural schematic diagram of a parity check disk array in the prior art; Figure 3 It is a structural diagram of the dual-control architecture of the prior art; Figure 4 A schematic diagram of a structure in which a dual-control architecture in the prior art uses memory as a cache; Figure 5 It is a flow chart of the high-speed centralized storage active-active method of the present invention; Figure 6 A flowchart of executing data writing in the high-speed centralized storage active-active method of the present invention; Figure 7 A flow chart of executing data reading in the high-speed centralized storage active-active method of the present invention; Figure 8 It is a structural schematic diagram of the high-speed centralized storage active-active device of the present invention. DETAILED DESCRIPTION
[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0019] It should be noted that when a component is referred to as being "connected" to another component, it may be directly connected to the other component or there may be an intermediate component. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the technical field of this application. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.
[0020] Embodiment 1: like Figure 5-7 As shown, a high-speed centralized storage active-active method is applied to a multi-controller architecture, where the multi-controller architecture includes multiple controllers. The high-speed centralized storage active-active method includes the following steps: S1. Use the mirror disk array as a cache module for the parity disk array. The cache module is used to temporarily store data when at least two controllers of the multi-control architecture write data to the parity disk array at the same time. The parity disk array is used to store data. Both the mirror disk array and the parity disk array include several disks.
[0021] In one embodiment, the cache module is also used to temporarily store the data when the length of the written data is less than a preset threshold, and when the length of the written data is less than the preset threshold, the management module returns the physical address of the corresponding data read and write module belonging to the mirror disk array, otherwise, the management module returns the physical address of the corresponding data read and write module belonging to the parity disk array.
[0022] In one embodiment, the disk is a dual-port disk or a disk cabinet that can be shared by multiple controllers. Using a dual-port disk or a disk cabinet that can be shared by multiple hosts and other disk devices ensures that the disk is readable and writable to all controllers. The disk cabinet that can be shared by multiple controllers is, for example, a JBOF disk cabinet.
[0023] Among them, when multiple controllers write simultaneously, a free physical address space is allocated from the mirror disk array; otherwise, for data with a length less than a first preset threshold, a free physical address space is allocated from the mirror disk array, and for data with a length greater than or equal to the first preset threshold, a free physical address space is allocated from the parity disk array.
[0024] S2. One of the controllers in the multi-controller architecture is used as the main controller, and the other controllers are used as standby controllers. Data read / write modules and management modules are deployed on each controller. Only the main controller runs the data read / write modules and management modules at the same time, and the standby controller only runs the data read / write module. The data reading and writing module is used to receive the data reading and writing request from the host and execute the data reading and writing task accordingly; The management module is used for managing the cache module and the address space. The address space is used to represent the correspondence between the logical address and the physical address of the data.
[0025] In one embodiment, the management module is also used to move data from the cache module to the parity disk array when the cache space of the cache module is insufficient, and update metadata after the migration is completed, and save the updated metadata in the mirror disk array and the parity disk array.
[0026] The management module of the present invention (corresponding to Figure 6 , Figure 7 When managing the address space, the cache and space management module will cache small data write requests in the mirror disk array first. When the cached data reaches the preset amount, the data will be merged into a large data and written to the parity disk array, thereby solving the write amplification problem caused by writing small data in the parity disk array and making full use of the high-speed bandwidth of the mirror disk array. The IOPS (Input / Output Operations Per Second, i.e. the number of read and write (I / O) operations per second) of writing small data can be increased by several times to several hundred times, which is a significant improvement.
[0027] In one embodiment, the management of the cache module adopts the bcache cache model or the ocf cache model; the management of the address space adopts the FTL model. The bcache cache model is the block device cache, the ocf cache model is the Open CAS Framework, and the FTL model is the flash translation layer. It is easy to understand that the management of the cache module and the address space can also adopt other existing technical models well known to those skilled in the art.
[0028] The software functions of the present invention are mainly divided into a management module and a data reading and writing module.
[0029] Data reading and writing module: The data read and write module runs on both the main controller and the standby controller, and is used to receive data requests from the host and perform data read and write tasks. When the data read and write module executes data read and write requests, it cannot directly read and write data according to the request address. It needs to use the read and write request address (logical address) as input to query the management module for the actual read and write address (actual storage address, i.e., physical address), and read and write data based on the actual read and write address.
[0030] Management Module: The management module only runs on the main controller and is responsible for the management of the cache module and the address space. The cache module is mainly used for temporary storage of write data to achieve device acceleration. The management of the address space is mainly used for the correspondence between the request address of the data and the actual storage address, in order to prevent conflicts. The management of the cache module adopts the bcache cache model or the ocf cache model; the management of the address space adopts the FTL model.
[0031] Specifically, the management module works as follows: 1) Responsible for the conversion between logical address and physical address. The logical address is the address exposed to the outside world, and the physical address is the actual storage address. 2) Responsible for communicating with the data read-write module, such as the data read-write module gets the logical address from the user, and tells the data read-write module where the physical address corresponding to the logical address is through the management module. 3) Allocate physical address space according to the logical address and length of the written data, such as allocating a free physical address space from the mirror disk array for data with a length less than the first preset threshold in step S1, which plays a role in cache acceleration. When multiple controllers write at the same time, allocate a free physical address space from the mirror disk array to prevent competition conflicts, thereby solving the problem that multiple controllers cannot write to the parity disk array at the same time, that is, when multiple controllers write at the same time, the management module returns the physical address of the data read-write module to the mirror disk array, and the mirror disk array can be written by multiple controllers at the same time. 4) It is responsible for rewriting the data of the mirror disk array to the parity disk array. The mirror disk array is only used as a high-speed data temporary storage area for caching data and resolving conflicts (the mirror disk array wastes space), and is not used for storing large amounts of data. It is easy to understand that the management module can also adaptively adjust functions, such as managing the storage space of the parity disk array and integrating fragmented storage space. These technologies are well known to those skilled in the art and will not be described in detail here.
[0032] In one embodiment, the multi-controller architecture is a dual-controller architecture, a quad-controller architecture, or an eight-controller architecture.
[0033] Among them, the present invention can be flexibly expanded and applied to multi-controller environments (such as dual-controller architecture, quad-controller architecture, eight-controller architecture, etc.). The required prerequisite is that the selected disk or disk cabinet can be shared by all controllers. In a multi-controller environment, the only controller running the management module is the main controller, and the other controllers are all backup controllers. The data read and write module runs in all controllers and is responsible for reading and writing data.
[0034] S3, using the data reading and writing module and the management module to execute the data reading and writing tasks according to the data reading and writing requests sent by the host, as follows: Data writing: Each data read-write module receives a data write request sent from the host end, and notifies the management module that there is data to be written to the disk, and requests the physical address space from the management module. The management module allocates the physical address space according to the logical address and length of the written data, and returns the physical address corresponding to the physical address space to the corresponding data read-write module. Each data read-write module writes the data to the corresponding physical address space according to the returned physical address. After the writing is completed, the management module is notified to update the metadata and save the updated metadata in the mirror disk array and the parity disk array. After the metadata is updated, the management module notifies the corresponding data read-write module. After receiving the notification, the data read-write module returns to the host end that the data write request has been completed, and returns the physical address corresponding to the physical address space to the corresponding data read-write module. Specifically: when there is simultaneous writing by the controller, the management module returns that the physical address of the corresponding data read-write module belongs to the mirror disk array, otherwise, the management module returns that the physical address of the corresponding data read-write module belongs to the parity disk array; Data read: Each data read / write module receives a data read request from the host, and provides the logical address and length of the data to the management module. The logical address of the data is used as the input of the management module, and the physical address of the data is obtained from the management module. The data is read from the corresponding mirror disk array or parity disk array according to the physical address. After the read is completed, the management module is notified to update the metadata and return to the host.
[0035] For data writing, the process is as follows Figure 6, the data read and write module in the figure is responsible for receiving the data dropped from the host side. The data read and write module notifies the management module that there is data to be written to the disk. The management module allocates the actual disk address space (physical address space) as needed, and returns the physical address corresponding to the physical address space to the data read and write module. The data read and write module writes the data to the actual disk address space. After writing is completed, the management module is notified to update the metadata. For example, the metadata includes the L2P table (logical address to physical address table) and the P2L table (physical address to logical address table), which are used to record the correspondence between each logical address and the physical address. The metadata can be saved in the mirror disk array and the parity disk array, such as saving the L2P table in the metadata in the mirror disk array and the P2L table in the parity disk array, or allocating and saving the metadata according to actual needs. Furthermore, after the metadata changes, the modified part needs to be rewritten to the corresponding disk array for storage. When writing data, the management module returns the physical address space according to the allocation rule, and then modifies the table to modify the physical address corresponding to the logical address to the newly allocated physical address. When reading data, the management module queries the table and returns the physical address corresponding to the logical address. Among them, when the data reading and writing module starts to write data, the management module as the core needs to decide where to write the data. Depending on the selected cache model (bcache cache model or ocf cache model), there are certain differences in the method, but the strategy that must be followed is: if writing data directly to the parity disk array will cause competition conflicts or the length of the data is less than the first preset threshold, the data must be written to the cache module (i.e., the mirror disk array) for temporary storage. The subsequent management module will move the data from the cache module to the parity disk array at an appropriate time (such as when the cache space of the cache module is insufficient).
[0036] For data reading, the process is as follows Figure 7 In the figure, the data read / write module is responsible for receiving the data read request from the host. The data read / write module uses the data request address as input and the actual storage address obtained from the management module to read the data. After the read is completed, the management module is notified to update the metadata and return to the host.
[0037] The correspondence between the logical address and the physical address of the data can be continuously modified and adjusted under actual circumstances. For example, when writing data, the data read-write module provides the logical address to the management module, and the management module returns an idle physical address that will not cause competition conflicts, thereby establishing a correspondence between the two. When the management module moves data from the mirror disk array to the parity disk array, and when organizing the fragmented space of the parity disk array, it will move data from one physical address to another, which will involve modifying this correspondence.
[0038] A controller is generally a computer or server. When multiple controllers write to a parity disk array at the same time, conflicts will occur. As described in the background technology, a stripe of a parity disk array can only be written by one controller. When multiple controllers write, errors will occur in calculating the check block of the stripe. The parity disk array cannot support multiple controllers writing at the same time, because the writing of any data block of the parity disk array requires other data blocks of the same stripe to participate in calculating the check data block. If multiple controllers write different data blocks of the same stripe at the same time, there will be a data check calculation conflict problem. Therefore, in principle, the physical address space returned by the management module cannot cause any conflict between different controllers, so if the returned address space belongs to the mirror disk array, there will definitely be no conflict. If the returned space belongs to the parity disk array, the physical address space of the same stripe cannot be provided to another controller before the request is completed, otherwise a conflict will occur. By using the cache module as a high-performance cache space and the management module to solve the problem that the parity disk array cannot be written simultaneously in a multi-control environment, the active-active storage is better realized.
[0039] S4. When the main controller crashes or is abnormal, one of the backup controllers starts the management module to take over the work of the main controller; when the backup controller crashes or is abnormal, the main controller works normally; when there is an abnormality in the link between the controller and the host, no operation is required, and other controllers work normally.
[0040] In one embodiment, a standby controller starts a management module to take over the work of the main controller, specifically: The standby controller loads and restores the metadata stored in the mirror disk array and the parity disk array to the started management module and then restarts the operation.
[0041] Specifically, the takeover is as follows: among all the functions of the management module, the main thing that needs to be saved is the correspondence between the logical address and the physical address, and the relevant metadata is saved in the corresponding disk array. The takeover process is that the standby controller loads, restores, and finally reruns the metadata saved in the disk array. That is, the standby controller takes over the frozen or abnormal main controller as the new main controller and runs the management module and the data read and write module at the same time, and other standby controllers still only run the data read and write module.
[0042] The present invention divides the controller into a main controller and a backup controller from the software point of view. Both the main controller and the backup controller can perform data read and write services at the same time, but the main controller needs to run an additional management module to perform management services. The mirror disk array will serve as a cache module for the parity disk array to cache data, accelerate data reading and writing, and prevent conflicts caused by writing to the parity disk array at the same time. The parity disk array is used as a large-capacity data storage module to store most user data.
[0043] Furthermore, different configurations can be made for different volumes. For example, in a dual-controller architecture, for volume X, controller A is selected as the primary controller and controller B is selected as the backup controller. For volume Y, controller B is selected as the primary controller and controller A is selected as the backup controller. Evenly distributing the primary and backup controller configurations for different volumes is beneficial to load balancing among multiple controllers.
[0044] When the standby controller crashes or is abnormal, the main controller will not be affected and no operation is required; when the main controller crashes or is abnormal, the standby controller needs to start the management module to take over the work of the main controller, such as the standby controller taking over the L2P table and P2L table, and restarting the management module. The standby controller will continue to serve the host as the new main controller; when there is a link abnormality between the controller and the host, no operation is required, and other controllers can provide data services normally.
[0045] The specific implementation process is as follows: by Figure 6 For example, the data writing implementation process: (1) The data read / write module receives the data write request from the host and starts writing data; (2) The data read / write module requests the physical address space from the management module and specifies the logical address and length of the data to be written. For example, if it notifies that there is data to be written, the logical address of the data is A and the length is X. (3) The management module returns the physical address space to decide where to write the data. In principle, the returned physical address space should not cause any conflict between different controllers. (4) The data reading and writing module writes the data according to the physical address corresponding to the returned physical address space (such as writing data to disk N, address M); (5) The data reading and writing module notifies the management module that the data has been written and has been written to the specified location; (6) The management module updates the metadata and changes the physical address corresponding to logical address A to disk N, address M; (7) The management module notifies the data reading and writing module that the metadata update has been completed; (8) The data read / write module returns the message to the host that the data write request has been completed, i.e., the data write is completed, and returns the status to the host.
[0046] by Figure 7 Take the data reading implementation process as an example: (1) The data read / write module receives the data read request from the host and starts reading data; (2) The data read / write module tells the management module the logical address and length of the read data. For example, if it notifies the management module that there is a data read, the logical address of the data is B and the length is Y. (3) The management module calculates and returns the corresponding actual storage address (e.g., reading from disk I, physical address J) to determine where to read the data; (4) The data read / write module reads data from disk I, physical address J, returns data to the host and notifies the management module after completion; (5) The management module updates the metadata generated by reading the data. At the same time, the data read and write module returns the completion of the data read request to the host and returns the host status.
[0047] The present invention has been successfully deployed on the Chaowei dual-control storage server and the Feiteng dual-control storage server, and has successfully used the dual-control dual-active mode implemented by the high-speed centralized storage dual-active method to provide external storage services. Both dual controllers have data traffic, and the dual-control performance is twice as high as that of using only a single controller.
[0048] It should be understood that although Figure 5 , Figure 6 , Figure 7 The steps in the flowchart are shown in sequence according to the arrows, but these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the figure may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.
[0049] Embodiment 2: like Figure 8 As shown, a high-speed centralized storage active-active device includes a host, a multi-controller architecture, a mirror disk array and a parity disk array. The multi-controller architecture includes multiple controllers. The mirror disk array and the parity disk array are connected to each controller. Each controller is also connected to the host, and metadata is exchanged between the controllers through a low-latency link channel. The high-speed centralized storage active-active device is used to execute any high-speed centralized storage active-active method mentioned in Example 1.
[0050] The high-speed centralized storage active-active device can be flexibly expanded and applied to multi-controller environments (such as dual-controller architecture, quad-controller architecture, and eight-controller architecture). The hardware architecture of this embodiment is as follows Figure 8As shown, it is a dual-controller architecture, controller A and controller B are the main controller and the backup controller respectively, and the mirror disk array is a high-speed mirror disk array. And the device uses dual-port disks or disk cabinets (such as JBOF disk cabinets) that can be shared by multiple controllers and other disk devices to ensure that the disk is readable and writable to all controllers. By using two disk arrays, the parity disk array is used for large-capacity data storage, and the mirror disk array is used for data caching. The low-latency link channel between the dual controllers (controller A and controller B) is used for metadata interaction between controllers. The high-speed centralized storage active-active device is used to execute any of the high-speed centralized storage active-active methods mentioned in the above embodiment 1.
[0051] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0052] The above-described embodiments only express the more specific and detailed embodiments described in this application, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of this application, which all belong to the protection scope of this application. Therefore, the protection scope of the patent application shall be based on the attached claims.
Claims
1. A high-speed centralized storage active-active method, applied to a multi-controller architecture, the multi-controller architecture comprising a plurality of controllers, characterized in that: The high-speed centralized storage active-active method comprises the following steps: S1. Using a mirror disk array as a cache module of a parity disk array, wherein the cache module is used to temporarily store data when at least two controllers of a multi-control architecture simultaneously write data to the parity disk array, wherein the parity disk array is used to store data, and both the mirror disk array and the parity disk array include a plurality of disks; S2. One of the controllers in the multi-controller architecture is used as the main controller, and the other controllers are used as standby controllers. Data read / write modules and management modules are deployed on each controller. Only the main controller runs the data read / write modules and management modules at the same time, and the standby controller only runs the data read / write module. The data reading and writing module is used to receive the data reading and writing request from the host and execute the data reading and writing task accordingly; The management module is used for managing the cache module and the address space, and the address space is used to represent the correspondence between the logical address and the physical address of the data; S3, using the data reading and writing module and the management module to execute the data reading and writing tasks according to the data reading and writing requests sent by the host, as follows: Data writing: Each data read-write module receives a data write request sent from the host end, and notifies the management module that there is data to be written to the disk, and requests the physical address space from the management module. The management module allocates the physical address space according to the logical address and length of the written data, and returns the physical address corresponding to the physical address space to the corresponding data read-write module. Each data read-write module writes the data to the corresponding physical address space according to the returned physical address. After the writing is completed, the management module is notified to update the metadata and save the updated metadata in the mirror disk array and the parity disk array. After the metadata is updated, the management module notifies the corresponding data read-write module. After receiving the notification, the data read-write module returns to the host end that the data write request has been completed. The specific method of returning the physical address corresponding to the physical address space to the corresponding data read-write module is: when there is a controller writing at the same time, the management module returns the physical address of the corresponding data read-write module to the mirror disk array, otherwise, the management module returns the physical address of the corresponding data read-write module to the parity disk array; Data read: Each data read / write module receives a data read request from the host, and provides the logical address and length of the data to the management module. The logical address of the data is used as the input of the management module, and the physical address of the data is obtained from the management module. The data is read from the corresponding mirror disk array or parity disk array according to the physical address. After the read is completed, the management module is notified to update the metadata and return to the host. S4. When the main controller crashes or is abnormal, one of the backup controllers starts the management module to take over the work of the main controller; when the backup controller crashes or is abnormal, the main controller works normally; when there is an abnormality in the link between the controller and the host, no operation is required, and other controllers work normally.
2. The high-speed centralized storage active-active method according to claim 1, characterized in that: The cache module is also used to temporarily store the data when the length of the written data is less than a preset threshold, and when the length of the written data is less than the preset threshold, the management module returns the physical address of the corresponding data read and write module belonging to the mirror disk array, otherwise, the management module returns the physical address of the corresponding data read and write module belonging to the parity disk array.
3. The high-speed centralized storage active-active method according to claim 1, characterized in that: The management module is also used to move data from the cache module to the parity disk array when the cache space of the cache module is insufficient, and update metadata after the migration is completed, and save the updated metadata in the mirror disk array and the parity disk array.
4. The high-speed centralized storage active-active method according to claim 1, characterized in that: The one standby controller starts the management module to take over the work of the main controller, specifically: The standby controller loads and restores the metadata stored in the mirror disk array and the parity disk array to the started management module and then restarts the operation.
5. The high-speed centralized storage active-active method according to claim 1, characterized in that: The management of the cache module adopts the bcache cache model or the ocf cache model; the management of the address space adopts the FTL model.
6. The high-speed centralized storage active-active method according to claim 1, characterized in that: The disk is a dual-port disk or a disk cabinet that can be shared by multiple controllers.
7. The high-speed centralized storage active-active method according to claim 1, characterized in that: The multi-controller architecture is a dual-controller architecture, a quad-controller architecture, or an eight-controller architecture.
8. A high-speed centralized storage active-active device, characterized in that: The high-speed centralized storage active-active device includes a host, a multi-controller architecture, a mirror disk array and a parity disk array. The multi-controller architecture includes multiple controllers. The mirror disk array and the parity disk array are connected to each of the controllers. Each of the controllers is also connected to the host, and metadata is exchanged between the controllers through a low-latency link channel. The high-speed centralized storage active-active device is used to execute the high-speed centralized storage active-active method as described in any one of claims 1-7.
Citation Information
Patent Citations
Double-control double-active storage control system and method
CN103092786A
Odd-even check-based RAID (redundant arrays of inexpensive disks) striped mirror data distribution method
CN104281499A
Multi-active multi-control storage system with dual RAID data protection
CN112596673A
Business data file storage method, device and system based on storage strategy
CN115905159A
Application programming interface for storing data
CN116775327A
