分布式存储中精确修复最小带宽再生码的性能研究
发布时间：2018-11-07 21:31
【摘要】：分布式存储系统为保证数据可靠性,需要对数据进行冗余存储来应对由于节点失效所带来的数据不可靠性.基于矩阵积构造的精确修复最小带宽再生码除了能够显著降低系统的存储冗余,而且编码的构造参数之间没有约束限制,还能够显著降低修复带宽的开销,具有广阔的应用前景.然而,基于此编码方案所设计的分布式存储系统的性能开销并没有得到充分的研究和分析.针对该编码在分布式存储系统中数据上传、修复、下载3个阶段,分别比较CPU使用率、文件大小、缓冲区大小以及有限域大小对上述3个阶段中运算速度的影响,发现通过对相关参数进行合理配置,可以使得基于相应编码方案的分布式存储系统能够获得良好的运行性能.
[Abstract]:In order to ensure data reliability, distributed storage system needs redundant storage to deal with the data unreliability caused by node failure. In addition to reducing the storage redundancy of the system, the exact repair minimum bandwidth regenerative code based on matrix product can significantly reduce the overhead of the repair bandwidth, and there are no constraints between the coded construction parameters. It has broad application prospect. However, the performance overhead of the distributed storage system based on this coding scheme has not been fully studied and analyzed. According to the three stages of data upload, repair and download in distributed storage system, the effects of CPU usage, file size, buffer size and finite field size on the operation speed are compared. It is found that the distributed storage system based on the corresponding coding scheme can achieve good performance by reasonably configuring the relevant parameters.
【作者单位】： 智能信息处理上海市重点实验室(复旦大学计算机科学技术学院);
【基金】：国家自然科学基金项目(61171074) 国家“八六三”高技术研究发展计划基金项目(2009AA01A348) 教育部新世纪优秀人才支持计划基金项目(NCET-11-0113)
【分类号】：TP333
[Abstract]:In order to ensure data reliability, distributed storage system needs redundant storage to deal with the data unreliability caused by node failure. In addition to reducing the storage redundancy of the system, the exact repair minimum bandwidth regenerative code based on matrix product can significantly reduce the overhead of the repair bandwidth, and there are no constraints between the coded construction parameters. It has broad application prospect. However, the performance overhead of the distributed storage system based on this coding scheme has not been fully studied and analyzed. According to the three stages of data upload, repair and download in distributed storage system, the effects of CPU usage, file size, buffer size and finite field size on the operation speed are compared. It is found that the distributed storage system based on the corresponding coding scheme can achieve good performance by reasonably configuring the relevant parameters.
【作者单位】： 智能信息处理上海市重点实验室(复旦大学计算机科学技术学院);
【基金】：国家自然科学基金项目(61171074) 国家“八六三”高技术研究发展计划基金项目(2009AA01A348) 教育部新世纪优秀人才支持计划基金项目(NCET-11-0113)
【分类号】：TP333
【相似文献】
相关期刊论文 前10条
1 孟庆伟;;云计算技术及其电信应用[J];电信快报;2010年07期
2 卢益阳;;分布式存储系统调查[J];企业科技与发展;2011年16期
3 王大涛;姜月秋;;一种适用于集群的高可用存储模型的研究[J];沈阳理工大学学报;2007年02期
4 邢继元;张义德;;计算机存储技术的发展现状和趋势[J];科协论坛(下半月);2008年02期
5 张天祥;潘峰;杨晓元;刘镇;张薇;;改进的LPCA算法[J];计算机工程与应用;2010年09期
6 陈卫卫;吴海佳;胥光辉;;分布式存储中文件分割的最优化模型[J];解放军理工大学学报(自然科学版);2010年04期
7 王文奎;吴国新;;一种对等式存储系统的设计与实现[J];计算机技术与发展;2008年04期
8 马军;郑p，
本文编号：2317623
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2317623.html
