Efficient RAID-6 MDS code for degraded reads optimization

Abstract

Abstract: With the coming of Big Data Era and the development of modern storage systems, there is a big challenge in the reliability of storage systems in modern data center. Due to the fact that most classic RAID-6 MDS codes have limitations on degraded reads, a new RAID-6 MDS erasure code for optimizing the degraded reads is proposed,which adjusts the diagonal layout of X-code to horizontal direction by taking advantage of horizontal codes while the continuous data elements are more likely to share the same horizontal parity chain for optimizing degraded reads. Experimental results show that the optimized X-code can efficiently reduce the additional I/O requests for degraded read operations. Especially, compared with RDP code and X-code, the optimized X-code can obtain an improved performance with the same number of disks.

Key words: RAID-6, MDS erasure codes, degraded reads

CLC Number:

V219
TP302.8

LAI Ran, FENG Xingjie, WANG Qing, LI Jie, JIAO Wenhuan. Efficient RAID-6 MDS code for degraded reads optimization[J]. .

References

[1] 智研咨询. 2016-2022 年中国大数据行业深度分析及投资战略咨询报告, R439946[R]. 北京: 智研集团, 2016.
[2] PINHEIRO E, WEBER W, BARROSO L A. Failure Trends in a Large Disk Drive Population[C]//5th USENIX Conference on File and Storage Technologies, San Jose, California, Feb 13-16, 2007: 1-20.
[3] BORTHAKUR D. The Hadoop Distributed File System: Architecture and Design[EB/OL]. http：//hadoop.apache.org/common/docs/current/hdfsdesign.html. (2017-02-21) [2018-02-10].
[4] LIN W K, CHIU DAH MING, JACK Y B. Erasure Code Replication Revisited[C]//4th International Conference on Peer-to-Peer Computing,Zurich, Ang 25-27, 2004: 90-97.
[5] WU CHENTAO, HE XUBIN, WU GUANYING, et al. HDP Code: A horizontal-Diagonal Parity Code to Optimize I/O Load Balance in RAID-6[C]//IEEE/IFIP 41st International Conference on Dependable Systems &Networks(DSN), Hongkong China, June 27-30, 2011: 209-220.
[6] FORD D,LABELLE F, POPOVICI F I, et al. Availability in globally distributed Storage Systems[C]//9th USENIX Symposium on Operating Systems Design and Implementation, Vancouver, BC, Canada, Oct 4 -6,2010: 61-74.
[7] 傅颖勋, 文士林, 马礼, 等. 纠删码存储系统单磁盘错误重构优化方法综述[J]. 计算机研究与发展, 2018, 55(1): 1-13.
[8] XIANG LIPING, XU YINLONG, JOHN C S LIU, et al. Optimal Recovery of Single Disk Failures in RDP Code Storage Systems[C]//ACM SIGMETRICS 2010, New York, June 14-18, 2010: 119-130.
[9] WANG ZHIYING, DIMAKIS A, BRUCK J. Rebuilding for Array Codes in Distributed Storage Systems[C]//2010 IEEE Global-Com Workshops,Miami, Florida, 2010: 1905-1909.
[10] XU SILEI, LI RUNHUI, LEE P, et al. Single disk failure recovery for Xcode-based parallelstoragesystems[J].IEEETransactionson Computers,2013, 63(4): 995-1007.
[11] XU LIHAO, BRUCK J. X-Code: MDS array codes with optimal encoding[J].IEEETransactionson Information Theory,1999,45(1):272-276.
[12] KHAN O, BURNS R, PLANK J, et al. Rethinking Erasure Codes for Cloud File Systems: Minimizing I/O for Recovery and Degraded Reads[C]//10th USENIX Conference on File and Storage Technologies. Berkeley, CA: USENIX Association, Jan, 2012: 1-14.
[13] WANG NENG, XU YINLONG, LI YONGKUN, et al. OI-RAID: A Two LayerRaidArchitectureTowardsFast RecoveryandHigh Reliability[C]//2016 IEEE Conference on Dependable Systems and Networks,Toulouse, France, June 28-July 1, 2016: 61-72.
[14] FU YINGXUN, SHU JIWU, SHEN ZHIRONG, et al. Reconsidering single disk failure recovery for erasure coded storage systems: Optimizing load balancing in stack-level[J]. IEEE Transactions on Parallel and Distributed Systems, 2017, 27(5): 1457-1469.
[15] FU YINGXUN, SHU JIWU, LUO XIANGHONG, et al. Short Code: An efficient RAID-6 MDS code for optimizing degraded reads and partial stripe writes[J]. IEEE Transactions on Computers, 2017, 66(1): 127-137.
[16] CORBETT P, ENGLISH B, GOEL A, et al. Row-Diagonal Parity for Double Disk Failure Correction[C]//3rd USENIX Conference on File and Storage Technologies, San Francisco, California, March 31-April 2,2004: 2-15.