Multi-node metadata management method for persistent memory file system in NUMA system

Abstract

The invention discloses a multi-node metadata management method for a persistent memory file system in an NUMA system, and relates to the technical field of novel storage of data centers. File data access is a mixing process of file metadata and file data, for example, in a file creating and deleting process, global information of a file system needs to be accessed so as to update access time, idle data blocks, idle inode and the like of the file system; in the file reading and writing process, a directory structure needs to be traversed, a file metadata inode structure corresponding to a file name is searched, space is distributed, data is copied, file system global information and file metadata inode information are updated, and file reading and writing are completed. In the design of a current NUMA persistent memory file system, the operation of remotely accessing metadata is involved in the file data access process, so that the file read-write performance is reduced. According to the method, a redundant dynamic attribute metadata management mechanism and a centralized directory management mechanism based on the access frequency are adopted, metadata management of the persistent memory file system in the NUMA system is optimized, and the influence of cross-node metadata access on file data access performance is reduced.

Description

Technical Field

[0001] The present invention relates to the field of novel storage technology for data centers, and in particular to a multi-node metadata management method for a persistent memory file system in a NUMA system. Background Art

[0002] The rapid expansion of Internet and IoT applications has led to an explosive growth of global data, which needs to be stored efficiently. Building a reliable and efficient data storage system has become an urgent problem to be solved. A new paradigm of heterogeneous storage based on hybrid DRAM and persistent memory (PM) has been designed to process data directly in or near the storage device to minimize data movement and improve performance, such as in-memory storage, in-memory computing, etc. Many persistent memory-based file systems are designed to take advantage of the byte-addressable, persistent and high-speed characteristics of persistent memory, such as EXT4-DAX, SCMFS, PMFS, NOVA, SIMFS, HiNFS. These persistent memory file systems use persistent memory as a data storage device and directly copy file data between the application user buffer and the persistent memory device, avoiding the software overhead of the traditional block device file system through the operating system cache and I / O software stack, and achieving better performance improvement.

[0003] In order to achieve greater throughput, reliability and economy of computer systems, the NUMA (Non-Uniform Memory Access) architecture is widely used in high-performance computing and cloud computing data center servers. The NUMA architecture system divides the processor and memory into node management, avoiding the scalability problem caused by the increasing memory access conflicts when multiple processors address the same memory address space, and has lower local memory access latency and better scalability. Many basic software have made special optimizations for NUMA architecture computer systems. For example, Linux kernel memory management supports non-uniform memory access models, and the scheduler's load balancing uses NUMA nodes as management domains. At the same time, many research works have proposed data allocation strategy mechanisms for NUMA architecture. For example, JeriFS redesigned the I / O stack to achieve optimal thread and buffer affinity. In addition, enterprises have made special treatments for virtualization platforms for NUMA architecture. For example, VMWare designed a NUMA-aware virtual machine I / O management scheduler to obtain maximum efficiency from local I / O devices.

[0004] Many studies have designed high-performance persistent memory file systems for NUMA architecture. NThread proposed a NUMA-aware thread migration module for persistent memory file systems, which designed a NUMA-aware thread migration module that does not require data migration. JHKim et al. virtualized persistent memory devices across NUMA nodes into a single address space and designed a local write-first memory allocation strategy to reduce remote memory access. HydraFS divides the file system storage space into multiple parts according to the number of NUMA nodes and schedules threads to the nodes where the files are located so that threads can get the best benefits from local file access.

[0005] The file system uses a directory hierarchical structure to manage file data. The file system contains file system global information and files. Files contain file metadata and file data. File system global information is used to describe the global characteristics of the file system, including file system size, file system read and write time, total number of file metadata inodes, total number of remaining file metadata inodes, total number of data blocks, number of allocated blocks, number of remaining data blocks, etc. File metadata is used to describe file characteristics, including owner, access time, permissions, file identifiers, etc. File data access is a mixed process of file metadata and file data. For example, in the process of creating and deleting files, it is necessary to access the file system global information to update the file system access time, free data blocks, free inodes, etc.; in the process of reading and writing files, it is necessary to traverse the directory structure, find the file metadata inode structure corresponding to the file name, allocate space, copy data, update the file system global information and file metadata inode information, and complete the reading and writing of files. In the current NUMA persistent memory file system design, the file data access process involves operations that remotely access metadata, resulting in a decrease in file read and write performance. How to efficiently manage metadata on multi-node space and reduce remote access to file metadata is an important challenge. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention proposes a multi-node metadata management method for a persistent memory file system in a NUMA system, aiming to optimize the metadata management of the persistent memory file system in the NUMA system and reduce the impact of cross-node metadata access on file data access performance.

[0007] To this end, the technical solution adopted by the present invention is a method for managing multi-node metadata of a persistent memory file system in a NUMA system, comprising:

[0008] (1) Redundant dynamic attribute metadata management based on access frequency: manage the file system global information and file metadata of the NUMA persistent memory file system, including

[0009] (1.1) Redundant dynamic attribute management mechanism for file system global information, managing file system global information on multiple NUMA nodes;

[0010] (1.2) File system global information redundant dynamic attribute synchronization mechanism, which synchronizes the dynamic attributes of file system information between multiple nodes through two synchronization mechanisms: active access synchronization and timed synchronization;

[0011] (1.3) File metadata redundant dynamic attribute management mechanism, which distributes all file metadata in the file system on a single node and manages the metadata attributes of all files through the index node table;

[0012] (1.4) File metadata redundant dynamic attribute synchronization mechanism, which synchronizes the file metadata dynamic attributes between the master node and the node where the file data is located through an active access synchronization mechanism;

[0013] (2) Centralized directory management: Manage the directory structure in the NUMA persistent memory file system, allocate space on a single node to construct the directory tree, and avoid access to remote persistent memory during directory retrieval.

[0014] Furthermore, in the step (1.1), all attributes in the file system global information are managed on node 0, and the file system global information attributes are divided into static attributes and dynamic attributes according to the access frequency of each attribute in the file system information. The dynamic attributes are redundantly stored at the starting position of each node, and the redundant file system global information dynamic attributes on the local node are accessed during the file access process to avoid the impact of remote file system global information access on file data access performance. In the file system global information structure of each node, a synchronization flag attribute is added.

[0015] Furthermore, step (1.2) adds a synchronization flag to the structure to indicate whether the field is modified. Whenever the file system global information is modified, the synchronization flag field is set. When there is a modification to the redundant dynamic attributes of the file system global information, it is first saved in the redundant backup of the local node. When the user actively accesses the file system global information and the synchronization cycle arrives, the modification order is synchronized to the file system global information of node 0 based on the synchronization flag and the modification time.

[0016] Furthermore, in order to quickly locate the file metadata inode of a specific file in step (1.3), the index node table is managed by an array, and the free index nodes in the array are linked by a linked list to achieve fast allocation and recovery; a space is reserved at the node where the file data is located for redundant storage of the metadata dynamic attributes of the file on the current node; during the file read and write access process, the reading and writing are completed through the redundant file metadata dynamic attributes on the local node to avoid the impact of remote file metadata access on the file data access performance. Specifically, the metadata attributes of all files are managed on node 0, and the file metadata attributes are divided into static attributes and dynamic attributes according to the access frequency of each attribute in the file metadata, and the dynamic attributes with high access frequency during the file access process are redundantly stored in the NUMA node where the file data is located.

[0017] The file metadata redundant dynamic attribute synchronization mechanism synchronizes the file metadata dynamic attributes between two nodes (a primary node and a node where the file data is located) through an active access synchronization mechanism.

[0018] Furthermore, a multi-node centralized directory tree management mechanism realizes centralized directory management in the NUMA persistent memory file system, avoiding the problem of cross-node access to directory files during vertical directory retrieval.

[0019] The centralized directory tree management mechanism stores the directory structure in the NUMA persistent memory file system on node 0; when creating a directory, the directory metadata structure and directory data block are allocated from node 0, and a tree-shaped directory hierarchical management structure is constructed on a single NUMA node, avoiding multiple remote access traversal of the directory between multiple nodes.

[0020] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned method for managing multi-node metadata of a persistent memory file system in a NUMA system are implemented.

[0021] The present invention also provides a system architecture of the management method of the present invention, including a file system global information redundant dynamic attribute management module, a file system global information redundant dynamic attribute synchronization module, a file metadata redundant dynamic attribute management module, a file metadata redundant dynamic attribute synchronization module and a multi-node centralized directory tree management module. In a NUMA architecture server, a processor, a DRAM and a persistent memory are divided into nodes for management. The nodes are connected through a high-speed interconnection bus. The persistent memories on multiple nodes are used as data storage devices for a persistent internalized memory file system. The persistent memory file system manages multi-node persistent memory space and stores file system global information, file metadata and file data in the persistent memory. The file system global information redundant dynamic attribute management module manages redundant file system global information attributes on multiple nodes, the file system global information redundant dynamic attribute synchronization module is used to synchronize dynamic attributes on multiple nodes, the file metadata redundant dynamic attribute management module is used to manage redundant file metadata dynamic attributes on multiple NUMA nodes, the file metadata redundant dynamic attribute synchronization module is used to synchronize dynamic attributes on node 0 and the node where the file is located, and the multi-node centralized directory tree management module is used to manage directories and files in the NUMA persistent memory file system.

[0022] The present invention proposes a multi-node metadata management method for a persistent memory file system in a NUMA system, which reduces the impact of cross-node metadata access on file data access performance and improves the parallelism of metadata operations in the NUMA persistent memory file system through a redundant dynamic attribute metadata management mechanism based on access frequency and a multi-node centralized directory tree management mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of the NUMA multi-node metadata management method architecture;

[0024] Figure 2 Schematic diagram of the redundant dynamic attribute management mechanism of the file system global information;

[0025] Figure 3 Workflow diagram of file system global information redundant dynamic attribute synchronization mechanism;

[0026] Figure 4 Schematic diagram of the redundant dynamic attribute management mechanism for file metadata;

[0027] Figure 5 Workflow diagram of file metadata redundant dynamic attribute synchronization mechanism;

[0028] Figure 6 Schematic diagram of multi-node centralized directory tree management mechanism. DETAILED DESCRIPTION

[0029] Figure 1 The schematic diagram of the NUMA multi-node metadata management method architecture includes a file system global information redundant dynamic attribute management module, a file system global information redundant dynamic attribute synchronization, a file metadata redundant dynamic attribute management module, a file metadata redundant dynamic attribute synchronization module, and a multi-node centralized directory tree management module. In a NUMA architecture server, processors, DRAM, and persistent memory are divided into node management, and the nodes are connected through a high-speed interconnect bus. The persistent memory on multiple nodes is used as a data storage device for a persistent internalized memory file system. The persistent memory file system manages the multi-node persistent memory space and stores the file system global information, file metadata, and file data in the persistent memory. The file system global information is used to describe the file system global characteristics, and the file system global information redundant dynamic attribute management mechanism manages the redundant file system global information attributes on multiple nodes. The file system global information redundant dynamic attribute synchronization mechanism synchronizes the dynamic attributes on multiple nodes. The file metadata is used to describe the file attribute characteristics, and the file metadata dynamic attribute redundant management mechanism manages the redundant file metadata dynamic attributes on multiple NUMA nodes. The file metadata redundant dynamic attribute synchronization mechanism synchronizes the dynamic attributes on node 0 and the node where the file is located. The multi-node centralized directory tree management mechanism manages directories and files in the NUMA persistent memory file system, allocates space on a single node to construct a directory tree, and avoids the access of remote persistent memory during the directory retrieval process.

[0030] Figure 2 The schematic diagram of the redundant dynamic attribute management mechanism of the file system global information is as follows. The file system global information is used to describe the global characteristics of the file system, including the total size of the file system, the read and write time of the file system, the data block size, the total number of file metadata structure inodes, the total number of remaining file metadata structure inodes, the total number of data blocks, the number of allocated blocks, the number of remaining data blocks, etc. According to the access frequency, these attributes are divided into static attributes and dynamic attributes. Static attributes remain unchanged during the access of the file system, such as the total size of the file system, the data block size, the Volume Name, the log space offset, the index node table offset, the mount time, etc. Dynamic attributes are constantly changing during the access of the file system, such as the file system write time, the number of allocated blocks, the number of remaining data blocks, the number of allocated inodes, the number of remaining inodes, etc. In addition, the file access process involves the operation of dynamic attributes. Based on the above analysis, the present invention is oriented to the NUMA multi-node architecture, and designs a redundant dynamic attribute management mechanism for the file system global information multi-node, and redundantly stores the dynamic attributes at the starting position of each node, such as Figure 2As shown. 512 bytes of space are reserved in the starting persistent memory space segment on each NUMA node to store the dynamic attributes of the file system global information. During file access, the dynamic attributes of the file system global information on the local node are accessed to avoid performance degradation caused by remote access to the file system global information. In the file system global information structure of each node, a synchronization flag attribute is added to the file system global information dynamic attribute synchronization mechanism on multiple nodes.

[0031] Figure 3 Workflow diagram of the synchronization mechanism for redundant dynamic attributes of file system global information. The dynamic attributes of file system global information have redundant backups on each node, and a synchronization mechanism needs to be designed. Considering the low frequency of user access to file system global information, the redundant dynamic attributes of multi-node file system global information are synchronized through active access synchronization and scheduled synchronization. In order to implement the synchronization mechanism for dynamic attributes of file system global information on multiple nodes, a synchronization flag is added to the structure to indicate whether the field is modified. Whenever the file system global information is modified, the synchronization flag field is set. The modification of the redundant dynamic attributes of the file system global information by file creation, deletion and other operations is first saved in the redundant backup of the local node. When the user actively accesses the file system global information and the synchronization cycle arrives, the modification order is synchronized to the file system global information of node 0 according to the synchronization flag and modification time. Figure 3 The workflow of the file system global information redundancy dynamic attribute synchronization mechanism is explained. Figure 3 (a) is a periodic synchronization mechanism:

[0032] In step S311, the system creates a periodic synchronization kernel thread when mounting the file system.

[0033] In step S312, the synchronous kernel thread determines whether the time interval period λ has arrived. If the time has not arrived, continue waiting, otherwise, execute S313.

[0034] In step S313, the synchronization kernel thread reads the synchronization flag in the dynamic attribute structure of the file system global information on each node. The setting of the synchronization flag indicates that the dynamic attribute on the node is updated.

[0035] In step S314, it is determined whether the node with the synchronization flag set is 0. If so, it means that synchronization is not required. Otherwise, the synchronization operation is performed in step S315.

[0036] In steps S315 and S316, the value on the master node is calculated and updated according to the data block number, inode number, etc. in the dynamic attribute set to 1. According to the time in the dynamic attribute set to 1, the latest time is found and the time on the master node is updated.

[0037] In step S317, it is determined whether the file system is unmounted, and if so, the kernel thread is terminated, otherwise it is continued to determine whether the synchronization period has arrived.

[0038] Figure 3 (b) For active access synchronization:

[0039] In step S321, the user invokes a VFS layer system call to access file system information.

[0040] In step S322, the NUMA file system receives a file system call request.

[0041] In step S323, the synchronization flag in the dynamic attribute structure of the file system global information on each node is read, and the setting of the synchronization flag indicates that the dynamic attribute on the node is updated.

[0042] In step S324, it is determined whether the node with the synchronization flag set is 0. If so, it means that synchronization is not required. Otherwise, the synchronization operation in S325 is executed.

[0043] In steps S325 and S326, the value on the master node is calculated and updated according to the data block number, inode number, etc. in the dynamic attribute set to 1. According to the time in the dynamic attribute set to 1, the latest time is found and the time on the master node is updated.

[0044] In step S327, the user request is returned based on the latest file system global information.

[0045] Figure 4 Schematic diagram of the redundant dynamic attribute management mechanism for file metadata. File metadata is used to describe the characteristics of a file, including owner, access time, permissions, file identifier, etc. Most file accesses require access to the file metadata first, and the frequency of access and modification of file metadata is much higher than that of file data. The efficiency of file metadata management has an important impact on the I / O performance of the file system. The various attributes of file metadata differ in access characteristics, size, and purpose. According to the access frequency, they are divided into static attributes and dynamic attributes. Based on the above analysis, the present invention redundantly stores the metadata dynamic attributes of the file on the current node at the node where the file data is located according to the access frequency of the file metadata attributes. Figure 4As shown in the figure, all file metadata in the file system are distributed on a single node, such as node 0, and managed through the index node table. In order to quickly locate the file metadata inode of a specific file, the index node table is managed by an array, and the free index nodes in the array are linked through a linked list to achieve fast allocation and recycling. A space is reserved at the node where the file data is located for redundant storage of the metadata dynamic attributes of the file on the current node. During the file read and write access process, the read and write are completed through the redundant file metadata dynamic attributes on the local node, avoiding the file data read and write frequently accessing the file metadata on the remote node.

[0046] Figure 5 Workflow diagram of the redundant dynamic attribute synchronization mechanism for file metadata. On the one hand, in actual applications, the frequency of file metadata access operations is lower than that of file read and write operations. On the other hand, in order to avoid remote access to file metadata during file access, the dynamic attributes of file metadata are redundantly backed up at node 0 and the node where the file data is located. Therefore, active file access synchronization is adopted between the two copies of the dynamic attributes of file metadata. After the file completes the read and write process, the hotspot attributes of the node where the file is located are actively synchronized to the master node to facilitate the operation of accessing file metadata in the later stage. Figure 5 The workflow of the file metadata redundant dynamic attribute synchronization mechanism is described as follows:

[0047] In step S501, the user calls the VFS layer system call to read and write file data.

[0048] In step S502, the NUMA file system receives a file system call request.

[0049] In step S503, the file read and write request operation is completed through the dynamic attributes of the file metadata on the local node.

[0050] In step S504, the updated file metadata dynamic attributes are remotely synchronized to the master node.

[0051] In step S505, the user file read and write request is returned.

[0052] Figure 6Schematic diagram of the multi-node centralized directory tree management mechanism. The multi-node centralized directory tree management mechanism manages the directory structure in the NUMA persistent memory file system, allocates space on a single node to construct the directory tree, and avoids the directory retrieval process from accessing the remote persistent memory. The directory tree structure is managed using a B-tree. When the file system is mounted, the directory inode and directory file data page are allocated on the primary node, such as node 0, to create the file system root directory. When the user calls the VFS system call to create a directory, the directory inode and directory file data page are allocated from the primary node to construct a B-tree directory structure. When the user creates a file, the file inode is allocated on the primary node, and the file metadata dynamic attribute structure is allocated on the target node by polling, and the dynamic attributes in the primary node file inode are copied to the metadata dynamic attribute structure on the target node. The purpose of separating directory retrieval and file metadata access from file data reading and writing is achieved.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: include: (1) Redundant dynamic attribute metadata management based on access frequency: manage the file system global information and file metadata of the NUMA persistent memory file system, including (1.1) Redundant dynamic attribute management mechanism for file system global information, managing file system global information on multiple NUMA nodes; (1.2) File system global information redundant dynamic attribute synchronization mechanism, which synchronizes the dynamic attributes of file system information between multiple nodes through two synchronization mechanisms: active access synchronization and timed synchronization; (1.3) File metadata redundant dynamic attribute management mechanism, which distributes all file metadata in the file system on a single node and manages the metadata attributes of all files through the index node table; (1.4) File metadata redundant dynamic attribute synchronization mechanism, which synchronizes the file metadata dynamic attributes between the master node and the node where the file data is located through an active access synchronization mechanism; (2) Centralized directory management: Manage the directory structure in the NUMA persistent memory file system, allocate space on a single node to construct the directory tree, and avoid access to remote persistent memory during directory retrieval.

2. According to claim 1, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: In the step (1.1), the file system global information attributes are divided into static attributes and dynamic attributes according to the access frequency of each attribute in the file system information, the dynamic attributes are redundantly stored at the starting position of each node, and the redundant file system global information dynamic attributes on the local node are accessed during the file access process to avoid the impact of remote file system global information access on file data access performance, and a synchronization flag attribute is added to the file system global information structure of each node.

3. According to claim 2, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: The static attributes remain unchanged during the access of the file system, including the total system size, data block size, Volume Name, log space offset, index node table offset, and mount time; the dynamic attributes change continuously during the access of the file system, including the file system write time, the number of allocated blocks, the number of remaining data blocks, the number of allocated inodes, and the number of remaining inodes.

4. According to claim 1, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: Step (1.2) adds a synchronization flag to the structure to indicate whether the field is modified. Whenever the file system global information is modified, the synchronization flag field is set. When there is a modification to the redundant dynamic attributes of the file system global information, it is first saved in the redundant backup of the local node. When the user actively accesses the file system global information and the synchronization cycle arrives, it is synchronized to the file system global information of node 0 based on the synchronization flag and modification time modification order.

5. A method for managing multi-node metadata of a persistent memory file system in a NUMA system according to claim 1 or 4, characterized in that: The timing synchronization in step (1.2) includes: Step S311, when mounting the file system, the system creates a periodic synchronization kernel thread; Step S312, the synchronous kernel thread determines whether the time interval period λ has arrived. If the time has not arrived, continue waiting; otherwise, execute S313; Step S313, the synchronization kernel thread reads the synchronization flag in the dynamic attribute structure of the file system global information on each node. The setting of the synchronization flag indicates that the dynamic attribute on the node is updated; Step S314, determining whether the node with the synchronization flag set is 0, if yes, it means that synchronization is not required, otherwise, executing the synchronization operation in S315; Step S315, calculating and updating the value on the master node according to the values ​​of the number of data blocks and the number of inodes in the dynamic attributes set to 1; Step S316, according to the time in the dynamic attribute set to 1, find the latest time and update the time on the master node; Step S317, determine whether the file system is unmounted, if so, terminate the kernel thread, otherwise continue to determine whether the synchronization period has arrived; The active access synchronization includes: Step S321, the user calls the VFS layer system call to access the file system information; Step S322, the NUMA file system receives a file system call request; Step S323, reading the synchronization flag in the dynamic attribute structure of the file system global information on each node, and setting the synchronization flag to indicate that the dynamic attribute on the node has been updated; Step S324, determining whether the node with the synchronization flag set is 0, if yes, it means that synchronization is not required, otherwise, executing S325 synchronization operation; Step S325, calculating and updating the value on the master node according to the values ​​of the number of data blocks and the number of inodes in the dynamic attributes set to 1; Step S326, according to the time in the dynamic attribute set to 1, find the latest time and update the time on the master node; Step S327, returning the user request based on the latest file system global information.

6. According to claim 1, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: In order to quickly locate the file metadata inode of a specific file in step (1.3), the index node table is managed by an array, and the free index nodes in the array are linked by a linked list to achieve fast allocation and recycling; a space is reserved at the node where the file data is located for redundant storage of the metadata dynamic attributes of the file on the current node; during the file read and write access process, the reading and writing are completed through the redundant file metadata dynamic attributes on the local node.

7. According to claim 1, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: Step (1.4) specifically includes Step S501, the user calls the VFS layer system call to read and write file data; Step S502, the NUMA file system receives a file system call request; Step S503, completing the file read and write request operation through the file metadata dynamic attributes on the local node; Step S504, remotely synchronizing the updated file metadata dynamic attributes to the master node; Step S505: Return the user's file read and write request.

8. According to claim 1, a method for managing multi-node metadata of a persistent memory file system in a NUMA system, characterized in that: The directory tree is managed by a B-tree. When the file system is mounted, the directory inode and directory file data pages are allocated on the master node to create a file system root directory. When a user calls the VFS system call to create a directory, the directory inode and directory file data page are allocated from the master node to build a B-tree directory structure; when a user creates a file, the file inode is allocated on the master node, and the file metadata dynamic attribute structure is allocated on the target node through polling, and the dynamic attributes in the master node file inode are copied to the metadata dynamic attribute structure on the target node.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for managing multi-node metadata of a persistent memory file system in a NUMA system according to any one of claims 1 to 8 are implemented.

10. A system architecture applicable to the management method according to any one of claims 1 to 8, characterized in that: It includes a file system global information redundant dynamic attribute management module, a file system global information redundant dynamic attribute synchronization module, a file metadata redundant dynamic attribute management module, a file metadata redundant dynamic attribute synchronization module and a multi-node centralized directory tree management module. In a NUMA architecture server, the processor, DRAM and persistent memory are divided into node management. The nodes are connected through a high-speed interconnect bus. The persistent memory on multiple nodes is used as a data storage device for a persistent internalized memory file system. The persistent memory file system manages multi-node persistent memory space and stores the file system global information, file metadata and file data in the persistent memory. The file system global information redundant dynamic attribute management module manages redundant file system global information attributes on multiple nodes, the file system global information redundant dynamic attribute synchronization module is used to synchronize dynamic attributes on multiple nodes, the file metadata redundant dynamic attribute management module is used to manage redundant file metadata dynamic attributes on multiple NUMA nodes, the file metadata redundant dynamic attribute synchronization module is used to synchronize dynamic attributes on node 0 and the node where the file is located, and the multi-node centralized directory tree management module is used to manage directories and files in the NUMA persistent memory file system.

Images