基于分组码的跳跃纠删码  

Jump erasure code based on block code

作　　者：孟宇龙 关智允 徐东[1] 张子迎 任龙 MENG Yulong;GUAN Zhiyun;XU Dong;ZHANG Ziying;REN Long(College of Computer Science and Technology,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学计算机与科学技术学院

出　　处：《哈尔滨工程大学学报》2019年第11期1896-1902,共7页Journal of Harbin Engineering University

基　　金：装发重大预研项目(31511030201)

摘　　要：针对纠删码中分组码多组之间关联性差导致的容错率低等问题,本文基于分组码的思想提出一种跳跃纠删码—跳跃局部重构码。通过将初始数据分组并且在每组选取单块数据跳跃再生成校验块来加强组间联系,以提升容错率、容错能力以及降低重构开销等不同性能,并且可以权衡存储开销与其他性能以满足分布式系统的不同需求。此外,本文通过改变跳跃局部重构码自身参数来对比性能变化,并且与其他常用类型纠删码进行对比实验。结果表明:跳跃局部重构码能够在较小的存储开销下,达到较高的容错能力和较低的重构开销的效果,可跳跃生成校验块并且相对性能较优。Aiming at the problem of low fault tolerance rate caused by poor correlation between multiple groups of block codes in the erasure code,this paper proposes a jump erasure code called jump local reconstruction code(JLRC)based on the idea of block code.Different performances such as fault tolerance rate,fault tolerance capability,and reconstruction overhead could be improved by grouping the initial data and selecting a single block of data in each group to generate parity blocks.Moreover,JLRC can also balance the storage overhead with other performance to meet different needs of distributed systems.In this study,by varying the parameters of the JLRC code,the performances at different parameters were compared,and the JLRC was compared with other commonly used types of erasure codes.The results show that the JLRC can achieve higher fault tolerance and lower reconstruction overhead with less storage overhead and can,by jumping,generate check blocks with relatively better performance.

关 键 词：纠删码 分组码 跳跃局部重构码 存储开销 容错能力 容错率 重构开销 

分 类 号：TP309.3[自动化与计算机技术—计算机系统结构]