Method of improving disk array data reconstruction efficiency

Abstract

The invention provides a method of improving the disk array data reconstruction efficiency. The method is characterized in that a reconstruction process of treating a same stripe is carried out in two threads, as for one thread, disk reading operation is carried out only, as for the other thread, disk writing operation is carried out, and the amount of data blocks of a buffer queue is limited and is larger than that of stripe buffer so as to asynchronously write data without influencing the disk reading operation; the data of an original data disk is read and is checked, and the data obtained by checking is added into the buffer queue; and the other thread used for disk writing acquires the data from the buffer queue. With the adoption of the method, the position of writing the data per time needs to be marked and recorded and can be updated into super block information, thus preventing reconstruction starting anew after restart due to power off.

Description

technical field [0001] The invention relates to the technical field of computer systems and storage, in particular to a method for improving data reconstruction efficiency of disk arrays. Background technique [0002] At present, soft RAID in the kernel reconstructs data according to stripes, reads all data disks and writes the obtained data to the added hot spare disk after verification; this process must include read and write operations, divided into two parts step is completed, read and write operations have to wait for each other, which is a serial operation, which reduces efficiency (such as figure 1 ). [0003] For a RAID5 system with redundant verification information, the data of each member disk cannot guarantee the consistency of the stripe data when the RAID5 is newly created. The inconsistency of the data makes the system unable to process the host IO request according to the normal process, and the obtained verification Data may be wrong, so these levels of R...

AI Technical Summary

Problems solved by technology

RAID5 supports the failure of one disk. When a disk fails and degrades, adding a hot spare disk for recovery is equivalent to performing a consistency check. The task must traverse all the data on the member disk, which is a very important task. Time-consuming work, and as the capacity of the disk increases, the process will take longer

[0004] In this process, if another disk fails, due to the redundancy mechanism of RAID5, it cannot be recovered, and the entire array will fail.

Images