A Synchronous Write-Back Method of Disk Data

Abstract

The present invention discloses a synchronized writeback method for disk data. The method comprises: setting a write cache space, comparing and acquiring difference calculation results diff of data blocks A and A' that have the same start sector number, and substituting the data block A for the data block A'; assigning a weight value to each difference calculation result diff, and sorting the difference calculation results diff from largest to smallest; when a common data block B requests to be written, controlling to perform compression on the data block B, so as to save partial space, taking and inserting the difference calculation result diff with the largest weight into the partial space, and writing a newly-formed data block B into a Z-cluster whose first block is empty of a disk, and caching the data block B into the write cache space; and when a utilization rate of a write cache reaches a preset threshold, substituting a data block in the write cache for the data block on the disk, and changing the Z-cluster into an N cluster. The method improves an I / O bandwidth utilization rate of the disk, and effectively avoids write amplification of small data.

Description

technical field [0001] The invention relates to the technical field of computer data reading and writing, in particular to a write-back I / O scheduling system based on synchronous writing. Background technique [0002] The smallest logical addressable unit of a file system is a block, and the size of a block is generally 4KB. The smallest addressable unit in a block device is a sector. The sector size of most block devices is 512B, and a block consists of 8 sectors. Physically, a disk consists of many sectors, and logically, a disk consists of many block devices. [0003] The I / O scheduler is a layer of software between the file system and the block storage device. The application creates I / O requests by calling the read / write method or the mmap method, and these requests can be modifications to any location on any file. The file system neither simply submits requests in the order they are received, nor does it store requests to disk as soon as they are received. Instead,...

AI Technical Summary

Problems solved by technology

The reading and writing of the disk is based on the block, usually the size of the block is 4KB. For asynchronous writing, the system can accumulate the modification to a certain extent before writing to the disk. For synchronous writing, write several bytes to the file system. When the data is small, the file system must write 4KB data to the disk, which causes the write amplification of small data, wastes the I / O bandwidth of the disk, and affects the performance of the system

Images