摘要
S-RAID 5是一种适于顺序数据访问的节能磁盘阵列,为了提高其性能并保持节能效率,提出一种能量管理算法:逻辑空间预测法(LSF),该算法对S-RAID 5的I/O请求地址进行动态聚类分析,以获得I/O请求在逻辑地址空间内的分布区,然后求出各分布区的动态特性,并结合S-RAID 5的特殊数据布局,预测磁盘的工作状态并根据预测状态调度磁盘.实验表明,在节能效果相当的情况下,与典型算法TPM、Markov相比,LSF可有效消除SRAID 5的响应时间延迟.
S-RAID 5 is an energy-efficient RAID for sequential data access, an energy management algorithm, logical space forecast (LSF) algorithm, for S-RAID 5 was proposed to improve the performance of S-RAID 5 while maintains its energy efficiency. LSF performed dynamical cluster analysis on the addresses of the I/O requests, and obtained the areas of I/Os distributed in the logical address space. According the dynamic characteristics of above areas and special data layout of S-RAID 5, LSF forecasted the states of the disks in S-RAID 5, and scheduled the disks based on the predicted states. Experimental result shows that compared with classical algorithms TPM and Markov, LSF can effectively eliminate the response time delay of S- RAID 5 while gains similar energy saving.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2014年第2期166-170,共5页
Transactions of Beijing Institute of Technology
基金
国家"八六三"计划项目(2013AA01A212)