性能优化的双盘非同时故障容错布局分析

The Analysis of Performance-Optimization Layout Against Non-Simultaneous Double Disk Failures

在分析双盘非同时故障容错模型的基础上,本文建议两种性能优化的布局:预留和约束的校验散布容错布局。当出现单盘故障后,这两种布局在不需要立即替换故障磁盘的情况下自动过渡到另一个单磁盘故障容错的数据布局。约束的校验散布单盘故障布局便于提高校验写和单盘故障恢复性能,而且过渡算法实现较快。

Based on analyzing the model of non-simultaneous double disk failure tolerance, two performance-optimization layouts of reservation and constraint parity de-clustering are proposed. When a single disk fails, one layout does not need to replace the failed disk at once and automatically turn to another against a single disk failure. Constraint parity de-clustering layouts are convenient to improve the performance of parity writes and single disk failovers, and the transient algorithm can be implemented more quickly.