摘要
通过讨论由数据条纹、其镜像和奇偶校验组成的校验条纹的容许两个盘故障的阵列布局,给出了一种新的阵列布局策略:把校验条纹的单元划分作为阵列布局的一个主要步骤,使得容许两个故障的条件通过校验条纹的单元划分来实现。为讨论容许多个盘故障的阵列布局提供了一条新途径。给出了校验条纹的容许两个故障的单元划分条件以及多种划分。对每个l(m+2l2m+1),利用给出的l—划分,设计了一种容许两个盘故障的阵列布局方案。给出的阵列布局方案具有高的可靠性、高的吞吐量、好的I/O性能以及简单的编码和解码算法。
A new data layout strategy for tolerating two disk failures in disk array be presented by considering the placement schemes of data stripe,it's mirror and parity.The condition of tolerating two disk failures is achieved by us-ing the check stripe unit partition which is the main step of our new strategy.The conditions and schemes of check stripe unit partition for tolerating two disk failures are given.For every l(m+2l2m+1)and the l-partition,the authors present a new scheme for tolerating two disk failures in disk array.The new schemes provide high reliabiliy,high throughput ,excellent I /O performance and simple and efficient encoding and decoding algorithms.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第18期82-85,共4页
Computer Engineering and Applications
关键词
镜像
奇偶校验
故障
磁盘阵列
数据布局
不可重构单元子集
Disk array,check stripe,check stripe layout ,check stripe unit partition,Unrecoverable unit subset