（國防科學技術大學 計算機學院，長沙 410073）
A novel way to repair data based on normal read bandwidth
ZHENG Han+,CHENG Hai-Tao, LU Yu-Tong
(Department of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, 410073, China)
Abstract:Cloud storage systems typically usereplication to providereliability. Recently, erasure codeshave been used to reduce the large storage overhead of three-replicated systems. Reed-Solomon codes are the standarddesign choice. But their high repair cost is often consideredan unavoidable price to pay for high storage efficiency andhigh reliability. This paper presents a new scheme based on normal read bandwidth to repair data to reduce the repair cost. We implement our algorithm in an open source storage system and make some test on it. The result shows this new way saved almost half repair cost compared with previous way.
Key words: data repair, cloud storage, erasure codes, repair cost optimization
摘要：云存儲系統主流采用副本的方式提供數據可靠性保證。當前，糾刪碼的方式正逐漸被采用取代傳統的三副本方式，以節省存儲開銷。業界目前主要選擇Reed-Solomoncodes作為糾刪碼的實現方式，但是，Reed-Solomoncodes較高的修復開銷對于高效率和高可用性的存儲系統而言往往是不可避免的。本文針對這一問題，提出一種基于正常讀帶寬的數據修復方法，減少數據修復過程中的開銷。我們在一個開源的存儲系統上實現了我們的算法，并進行測試比較，與之前的實現相比，要節約大約一半的修復開銷。
關鍵詞：數據修復，云存儲，糾刪碼，修復開銷優化
References:
[1]田敬, 代亞非. P2P 持久存儲研究[J]. 軟件學報(Journal of Software). 2007, 18 (6):1379–1399.
[2]Cheng Huang et al.Erasure Coding in Windows Azure Storage. USENIX ATC’12, 2012
[3]Andrew Fikes. Storage Architecture and Challenges. In Faculty Summit, 2010.
[4]沈志榮,易樂天,舒繼武.大規模數據中心的數據存儲可靠性.國計算機學會通訊, 2012,8(10):8−16.
[5]Dabek F et al. Wide-Area cooperative storage with CFS.In Proc. of SOSP, 2001.
[6]黃震. 規模分布式存儲系統中數據冗余技術研究. 博士, 國防科學技術大學研究生院, 2012.
[7]Sit Eet al. Proactive replication for data durability [C]. InInternational Workshop on Peer-to-Peer Systems (IPTPS). 2006.
[8]Duminuco A et al. Proactive Replication in Distributed StorageSystems Using Machine Availability Estimation [C]. In Proc. CoNEXT’07.December 2007.
[9]Osama Khan et al.Rethinking Erasure Codes for Cloud File Systems:Minimizing I/O for Recovery and Degraded Reads.
[10]James S. Plank. Erasure Codes forStorage Applications.In Proc.Of the 4th USENIX Conference on FAST, 2005.
[11]James S. Plank et al. Optimizing Cauchy Reed-Solomon Codesfor Fault-Tolerant Network Storage Applications. In Proc. of NCA, 2006
[12]Maheswaran et al. XORing Elephants: Novel Erasure Codes for Big Data. The 39th International Conference on Very Large Data Bases,August 26th 30th2013, Riva del Garda, Trento, Italy.
[13]Tahoe-LAFS project.https://tahoe-lafs.org.
[14]EnoThereska et al. Sierra: Practical Power-proportionality for Data Center Storage. EuroSys’11, 2011, Salzburg, Austria
[15]Plank J. A tutorial on reed-solomon coding for fault-tolerance in RAID-like systems. Software Practice and Experience, 1997,27(9):995-1012.
作者簡介：
鄭涵（1987-），男，碩士生，主要研究分布式存儲，數據生存性。
