Method and system for transaction write optimization framework of persistent memory file system

Abstract

The invention provides a system and method for a transaction write optimization framework of a persistent memory file system. The system comprises a module M1, a module M2 and a module M3; the moduleM1 comprises a write file input metadata module, the metadata module dynamically backs up metadata in the write file system and supports metadata fault recovery; the module M2 writes a file and inputting the file into the data module, and the data module ensures the consistency of data and recovers the consistency state when a fault occurs; and the module M3 judges the atomicity of file metadata and data operation after the metadata module and the data module are completed, and writes the latest metadata information and / or data information judged by the atomicity decision logic module back tothe metadata module and / or the data module. The invention provides a transaction optimization read-write method mainly for solving the consistency realization problem of a persistent memory file system.

Description

technical field [0001] The invention relates to the technical field of computer system structure, in particular to a method and system for a transaction write optimization framework oriented to a persistent memory file system. Background technique [0002] Consistency is one of the most important properties of transactions. It is also a great challenge for technicians to maintain the consistent state of data in the file system. Existing consistency methods, such as logs, perform two double-write operations of data in one transaction, so that either all actions occur or none occur. Persistent memory, a new type of storage device, has the characteristics of non-volatility, byte addressability, and high performance. Traditional consistency implementations cannot make full use of the byte addressable characteristics of storage devices. At the same time, due to the limitations of transactions, Limit double-write to the critical path. Therefore, it is necessary and positive to ...

AI Technical Summary

Problems solved by technology

[0006] PM: Persistent Memory, non-volatile memory or persistent memory, STTRAM, PCM, and RRAM all belong to PM, which is characterized by no loss of information after power failure, and byte addressing; but the number of reads and writes is limited, and read and write Unbalanced, depending on the material, the PM write time is about 4 to 10 times the read time, and the PM read time is 1 to 4 times slower than DRAM

Images