存储系统中的局部修复阵列码模型  

Local repairable array code model in storage systems

作　　者：洪铁原 唐聃[1,2] 熊攀 蔡红亮 曾琼 许源平 Hong Tieyuan;Tang Dan;Xiong Pan;Cai Hongliang;Zeng Qiong;Xu Yuanping(School of Software Engineering,Chengdu University of Information Technology,Chengdu 610225,China;Sichuan Province Enginee-ring Technology Research Center of Support Software of Informatization Application,Chengdu 610225,China)

机构地区：[1]成都信息工程大学软件工程学院,成都610225 [2]四川省信息化应用支撑软件工程技术研究中心,成都610225

出　　处：《计算机应用研究》2024年第1期193-199,共7页Application Research of Computers

基　　金：四川省科技厅重大专项资助项目(2022ZDZX0001);四川省科技计划资助项目(2022YFG0033,2022YFG0037)。

摘　　要：对于单容错和双容错的存储系统,在磁盘修复过程中发生的任何故障都可能引起数据丢失,导致修复失败,保证数据的修复效率对于存储系统的可靠性至关重要。RDP码在进行单盘故障修复时使用混合恢复算法能减少25%的读取总量,但是在进行双盘故障修复时需读取所有的元素。针对目前难以同时提升单双盘故障修复效率的问题,对RDP码进行拓展,提出了一种具有局部修复性质的阵列码模型——DRDP码。DRDP码在RDP码的基础上将部分数据列按水平线进行异或计算生成局部水平校验列,并将其参与到全局校验列的编码计算中,从而缩短了修复链,使其拥有局部修复的功能。通过理论分析,DRDP码拥有良好的编译码复杂度和更新效率,大幅节省了单盘故障修复读取开销,并对双盘故障修复读取开销进行了优化,同时能修复75%三盘故障的情况。实验结果表明,与RDP码、LRRDP码和RDP(p,3)码相比,DRDP码的编码时间可节省8.23%~32.89%、单盘故障修复时间可节省7.08%~35.01%、双盘故障修复时间可节省5.07%~29.26%。For single fault tolerance and double fault tolerance systems,any failures that occurs during the disk repair process may cause data loss,leading to repair failure.Ensuring the efficiency of data repair is essential to the reliability of storage systems.RDP code(row diagonal parity code)can reduce the total number of reads by 25%when using the hybrid recovery algorithm for single disk failure repair,but all elements need to be read when repairing double disk failure.To address the current difficulties in improving the efficiency of single and double disk failure repair at the same time,this paper extended RDP code and proposed an array code model with local repair properties,namely DRDP code(double rows diagonal parity code).Based on RDP code,DRDP code generated a local horizontal parity column by XOR calculation of some data columns according to horizontal lines and participated it in the encoding calculation of global parity columns,thus shortening the repair chain and making it have the function of local repair.Through theoretical analysis,DRDP code not only has good encoding and decoding complexity and update efficiency,but also significantly saves the single disk failure repair read overhead and optimizes the double disk failure repair read overhead.It can also repair 75%of three disk failure cases.The experimental results show that compared with RDP code,LRRDP code,and RDP(p,3)code,the encoding time of DRDP code can be saved by 8.23%to 32.89%,the repair time of single disk failure can be saved by 7.08%to 35.01%,and the repair time of double disk failure can be saved by 5.07%to 29.26%.

关 键 词：阵列码 RDP码 存储系统 局部修复 读取开销 

分 类 号：TP333.3[自动化与计算机技术—计算机系统结构]