基于V码的高可靠性磁盘阵列容错模型

Fault-tolerant Model for RAID of High-reliability Based on V Codes

在磁盘阵列模型中,关键是如何实现容许多个磁盘阵列故障使得系统性能达到最优。该文提出了一类新的纠双错编码――V码,使用该编码的磁盘阵列数据布局,阵列的盘数可以为偶数,校验信息均匀分散在阵列每个盘中,容许任意2个磁盘故障。与其它纠双码的磁盘阵列布局进行比较,当阵列盘数为偶数时,V码阵列布局具有最优性能,编译码复杂度、冗余率达到最低以及小写性能最优,利于解决磁盘阵列I/O问题。

Redundant array of inexpensive disks （RAID） in the best way to reduce the disparity between the processing speed of CPU and the ability of I/O system. This paper presents a class of new binary maximum distance separable （MDS） array codes called V codes. The V codes are of minimum column distance 3 and have a simple geometrical structure, namely, they can correct either single burst column error or double column erasure errors. In addition, the parity symbols are evenly distributed among all the columns, thus the bottleneck effects of repeated write operation are naturally overcome. The complexity of encoding and decoding, the small write performance of three types of codes, which are V codes, EVENODD codes, RS codes, are compared and analyzed. The performances of RAID system using V codes are optimal.