A reconstruction optimization method based on erasure code cache

Abstract

The invention discloses a reconfiguration optimization method based on erasure code caching. The reconfiguration optimization method based on erasure code caching comprises a failure data caching process and a cached data migration process; the failure data caching process comprises the steps of receiving a client request record from a write miss data node of a client, writing the client request record in a data area in an erasure code cache region which is composed of pre-allocated memories of survived nodes in a cluster, updating a check area in the erasure code cache region according to the original data in the data area in the erasure code cache region and the new written client request record, and returning the completion result of the write miss data node to the client; the cached data migration process comprises the steps of reading one client request record in the erasure code cache region, reading data from a position of which a data offset address is offset in a substitute node corresponding to a failure node, and writing new data corresponding to the client request record in a position of which a write data offset address is offset in the substitute node. The reconfiguration optimization method based on erasure code caching is capable of improving the reconstruction performance of the failure data nodes and also capable of reducing the response time of the client to write miss.

Description

technical field [0001] The invention belongs to the field of computer storage, and more specifically relates to a reconstruction optimization method based on an erasure code cache. Background technique [0002] The erasure code cluster storage connects multiple storage nodes according to the erasure code method to form a storage system with certain fault tolerance. The multiple storage nodes constituting the erasure code cluster storage can be logically regarded as a large storage pool, and each node works at the same time to improve I / O parallelism. Due to the use of erasure coding algorithm, in the case of a certain number of storage nodes failure, the data on the failure nodes can be decoded to achieve the purpose of fault tolerance. Generally, data recovery operations and client I / O interact with each other. On the one hand, during the recovery process of lost data, the read and write performance of the cluster storage will decrease; on the other hand, compared with th...

AI Technical Summary

Problems solved by technology

The write invalidation process in traditional erasure code storage clusters has the following problems: First, the process of writing invalid data nodes and the process of reconstructing invalid data are carried out at the same time, and the two processes compete with each other for resources such as network bandwidth, disk, and memory, and the reconstruction time As a result, the reliability of the system is reduced; in addition, the client's write request involves three operations: recovery of invalid data, writing of new data, and updating of verification data, so the response time of the client's write request will also increase

Images