High-performance extensible lightweight file system based on NVM

Abstract

The invention proposes a high-performance extensible lightweight file system based on an NVM. The system comprises superblocks, an index node table, a hash table, a segment table, metadata logs, datalogs, bitmaps and a file data space; index nodes are stored in the index node table, and each node stores necessary metadata information; segments are stored in the segment table, and each segment stores continuous region information organized according to byte units; a file system naming layer is organized by the global hash table, and each hash bucket is a linked list with nodes with same hash values; data of each file is managed by a file B+ tree based on the segments, and each segment serving as a leaf node of the file tree represents file data segment; the metadata logs and the data logsall comprise multiple log files; each bitmap represents the use condition of the corresponding data block in the file system; the file data space stores file data and is managed with a 4KB block as aunit.

Description

technical field [0001] The invention relates to data storage, in particular to an NVM-based high-performance, scalable and lightweight file system. Background technique [0002] New non-volatile storage NVMs, such as Phase change memory (PCM), Spin Transfer TorqueRAM (STT-RAM), read and write speeds close to memory, can communicate directly with the CPU for byte addressing access, and are non-volatile and power off After the data will not disappear. The emergence of these non-volatile storage technologies will completely change the existing storage system model, and provide a high degree of parallelism and extremely low latency (nanosecond level) for data access. [0003] Although the performance of non-volatile storage NVMs is superior, it is precisely because of its fast read and write speed and non-volatile characteristics that it brings challenges to the design of the file system on it, resulting in prominent file system overhead and loss of data consistency. guarantee...

AI Technical Summary

Problems solved by technology

But on NVM whose access speed is close to DRAM, the lookup overhead on VFS becomes significant

Allocating, initializing, and releasing these memory cache structures also brings unnecessary overhead

In addition, the virtual file system VFS uses some global locks to protect concurrent insertion and deletion, which brings concurrency bottlenecks on multi-core systems

[0016] (2) Using the traditional copy-on-write and log methods to provide data consistency and bring redundant copy overhead

However, NOVA cannot identify dirty pages, and will write back all memory-mapped data to NVM during synchronization, even if the data is only read but not modified

Images