Metadata management method based on nonvolatile memory file system

Abstract

The invention discloses a metadata management method based on a nonvolatile memory file system. The method comprises the steps that 1, file index nodes are organized into an index node region in an array mode, and each index node is provided with a fixed serial number; 2, according to a system page table format, a mapping table is built for the file index node region, and the highest pointer of the mapping table is stored into a super block of the file system; 3, each file index node obtains a corresponding virtual address of itself through an ino number and an initial virtual address of the index node region; when the file system is mounted, the operating system analyzes a first table item of the system page table corresponding to the first virtual address space; 4, the super block of the file system is checked, the highest pointer of the mapping table of the index node region is inserted into the first table item, and the mapping relation between the virtual address space and the physical address space in the index node region is built. The method has the advantages that the nonvolatile memory writing abrasion balance is ensured, and metadata can be rapidly accessed.

Description

technical field [0001] The invention belongs to the technical field of computer memory management, and in particular relates to a metadata management method of a file system considering non-volatile memory wear leveling. Background technique [0002] With the emergence of new non-volatile memory, the business community and academia have proposed to make non-volatile memory with high storage density, fast read and write speed, addressable by byte and anti-vibration as a persistent data storage device , called Storage Class Memory (SCM for short). [0003] To manage storage-class memory, various in-memory file systems have been devised. For example, PMFS designed by IBM, "System software for persistent memory," in Proc. 9th ACM Euro. Conf. Comput. Syst., pp 15:1—15:15, 2014, ISBN: 978-1-4503-2704-6 ("System Software for Persistent Memory", 9th ACM European Computer Systems Conference, Vol. 15, pp. 1-15, 2014, ISBN: 978-1-4503-2704-6). SIMFS independently developed by Chongq...

AI Technical Summary

Problems solved by technology

[0007] The first is a tree structure. In this way, the storage management of file index nodes is very flexible and easy to expand. However, this method has a slow search speed and a large storage space overhead.

For example: BPFS organizes inodes in a tree structure, which makes inode allocation, recovery, and physical space selection very flexible and has good scalability, but it also makes inode retrieval more complicated, querying inodes takes more time, and many intermediate nodes will be lost. cause a large storage space overhead

[0008] The second is an array structure, which stores all file index nodes in a fixed area, and the size of the file index node is a fixed value, so the address of the file index node can be directly located through the file index node number, and this method can quickly find the file of the file Index nodes, but the scalability is poor, it is necessary to fix the number of file index nodes, that is, the total number of files supported by the fixed system; and fixing an area is very easy to cause wear and tear

Because its metadata is fixed in this physical area, each write is in this physical address area, which is easy to cause wear and tear of this physical area, that is, it cannot achieve wear balance in this area

Images