摘要
为改进传统RAID5机制小写性能低下问题,提出了一种高性能的冗余数据读写机制RMB(RedundancyManagement for RAID5).RMB采用异步冗余校验方法APC(Asynchronous Parity Calculating),对现有RAID5更新算法的关键步骤进行了优化,将统一计算校验块的同步操作拆分成同步镜像写更新和异步冗余计算更新.该操作分解了数据与其校验数据更新的紧耦合现状,提升了写操作并发度;通过合并多个版本镜像块写操作减少异或计算操作的次数.通过实验验证表,RMB机制,可将小写性能大幅提高,峰值可达4倍以上.
To solve the problem that RAID5 performance ineffective for small writes,a new high performance read and write mechanism for redundancy data is brought up,which is called RMB.By Asynchronous Parity Calculating.the mechanism RMR optimized the write-update mechanism of tradition al RAID5 and decomposed the concentrated operations of calculating parity blocks into writing mirror and asynchronous calculating.The new algorithm accomplish the decomposition of the tight data,improve the concurrency state and decrease the count of calculate by the way of merge ring several versions.It shows from experiments that RMR will substantially improve the performance of small writes,the peak value will be over four times as before under the out-of-band mode.
出处
《微电子学与计算机》
CSCD
北大核心
2012年第7期112-115,共4页
Microelectronics & Computer
基金
山东省自然科学基金项目(Y2011A03)
关键词
磁盘冗余阵列
冗余数据
小写更新
异步冗余计算
RAID
redundancy data
update small writes
asynchronous parity calculating