电子学报 ›› 2019, Vol. 47 ›› Issue (9): 1987-1993.DOI: 10.3969/j.issn.0372-2112.2019.09.024

• 学术论文 • 上一篇    下一篇

一种能耗优化的RAID小写预读方法

孙志卓1, 黄洪2, 张全新3, 谭毓安3, 李元章3, 张小松4   

  1. 1. 德州学院计算机系, 山东德州 253000;
    2. 中国人民解放军 61516部队, 北京 100094;
    3. 北京理工大学计算机学院, 北京 100081;
    4. 唐山学院计算机科学与技术系, 河北唐山 063000
  • 收稿日期:2018-08-30 修回日期:2018-12-31 出版日期:2019-09-25 发布日期:2019-09-25
  • 通讯作者: 张小松
  • 作者简介:孙志卓.E-mail:sunzhizhuo@126.com;黄洪.E-mail:huanghongbj@163.com;张全新.E-mail:zhangqx@bit.edu.cn;谭毓安.E-mail:victortan@yeah.net;李元章.E-mail:popular@bit.edu.cn
  • 基金资助:
    国家自然科学基金(No.61370063);山东省自然科学基金(No.ZR2016FM37,No.ZR2017LC014,No.ZR2015FL006)

A Small Write Prefetching Scheme for Engery-Optimized RAID Sub-systems

SUN Zhi-zhuo1, HUANG Hong2, ZHANG Quan-xin3, TAN Yu-an3, LI Yuan-zhang3, ZHANG Xiao-song4   

  1. 1. College of Computer Science and Technology, Dezhou University, Dezhou, Shandong 253000, China;
    2. Chinese People's Liberation Army Unit 61516, Beijing 100094, China;
    3. School of Computer Science, Beijing Institute of Technology, Beijing 100081, China;
    4. Department of Computer Science and Technology, Tangshan University, Tangshan, Hebei 063000, China
  • Received:2018-08-30 Revised:2018-12-31 Online:2019-09-25 Published:2019-09-25

摘要: 视频监控、备份、归档等应用产生海量存储数据,导致存储能耗急剧增加.S-RAID采用局部并行数据布局,可显著降低该类应用的存储能耗.为使更多磁盘待机节能,S-RAID通常执行"小写"操作,写操作时会额外引入等量的读操作,会显著降低性能.现有预读机制主要发生在文件级,无法感知RAID级小写引发的读旧数据、旧校验数据等读操作,因此不会也无法预读该类数据.为此,提出一种面向S-RAID的RAID级小写预读算法,由小写操作触发并在RAID级执行预读,根据S-RAID的数据布局方式,大粒度异步预读小写需要的旧数据、旧校验数据,有效减少I/O数和寻道数,提高磁盘的利用率.该方法可显著提高S-RAID的写性能,并且不依赖于任何额外硬件,具有更高的可用性.

关键词: 海量存储, 节能, 局部并行, 盘阵, 数据布局

Abstract: The applications,such as video surveillance,backup,and archiving,generate massive storage data,which make the energy consumption of storage devices increase rapidly.S-RAID can reduce the storage energy consumption of above applications significantly.In order to make more disks standby for energy saving,S-RAID prefers to small write.However,performing small write requires additional and equal amount of read operations for generating parity data,and the write performance of S-RAID is decreased.The existing prefetching mechanism mainly implements at the file system level,and cannot sense the read operations generated by small write at RAID level,such as reading old data and reading old parity.So it can not prefetch these old data and old parity data.Therefore,a RAID-level prefetching algorithm for small write of S-RAID is proposed,which is triggered by small write operations and performs at RAID level.It asynchronously prefetches the old data and old parity data required by small writes in large granularity according to the data layout of S-RAID.The write performance of S-RAID can be improved by about 40% without any additional hardware cost.

Key words: massive storage, energy conservation, partial parallelism, RAID, data layout

中图分类号: