一类精确修复多个节点的简单再生码

A CLASS OF SIMPLE REGENERATING CODES CAPABLE OF EXACT MULTI-NODE REPAIR

海量数据环境下要求存储系统具有高扩展性、高可靠性和低成本等特点。大规模存储系统的节点因数目巨大而易频繁失效,为保证节点的可用性,系统会利用冗余数据对失效节点进行修复。作为一种新的容错技术,再生码可有效降低分布式存储系统中失效节点修复时需要的下载数据量。基于简单再生码,为分布式存储系统设计一种新的编码方式。它不仅可容忍多个节点同时出错并进行修复,而且编码形式简单并具有较高的码率。

Massive data environment requires the storage system with the characteristics such as high scalability,high reliability and low price,etc. However,the nodes in large-scale storage system will frequently failure due to too huge in number. In order to ensure the usability of nodes,the system will use redundancy data to repair the failure nodes. As a new fault-tolerant technology,regenerating code can effectively reduce the amount of the download data required when repairing the failure nodes in distributed storage system. In this paper,we design a new encoding mode for distributed storage system based on simple regenerating codes. This mode can not only tolerates the simultaneous errors of multiple nodes and repairs them,but also has simple encoding form and achieves higher code rate.