计算机科学 ›› 2026, Vol. 53 ›› Issue (6): 367-375.doi: 10.11896/jsjkx.260200107
孙婧, 王弈, 陈海燕
SUN Jing, WANG Yi, CHEN Haiyan
摘要: 在现代多可用区(Multi-AZ)云存储系统中,故障呈现出显著的层次化与相关性特征。传统的纠删码(如RS码)或扁平化喷泉码方案往往忽略了物理拓扑结构,导致在处理频发的局部故障时产生高昂的跨AZ修复带宽,且难以兼顾异构网络下的灵活性。针对这一问题,提出了一种面向多可用区环境的分层AZ-Fountain编码模型。该模型将喷泉码的无码率特性与多AZ故障域结构进行显式耦合,设计了双层编码架构:在AZ内部采用局部优化双模态度分布(AZ-OBMD),以实现零跨区流量的局部修复;在全局采用Raptor-link度分布(AZ-RLD),以保障AZ级灾难下的高效恢复。实验结果表明,与RS码、LRC及最新的AZ-Code相比,所提方法在局部单块故障下能够与LRC一样保持零跨AZ流量,且其参与解码的数据块数量与修复延迟更低;在AZ级灾难恢复中显著降低了修复延迟与传输开销,实现了可靠性与修复代价的系统性权衡。
中图分类号:
| [1]REED I S,SOLOMON G.Polynomial codes over certain finite fields[J].Journal of the Society for Industrial and Applied Mathematics,1960,6(8):300-304. [2]LUBY M.LT codes[C]//Proceedings of The 43rd Annual IEEE Symposium on Foundations of Computer Science.2002:271-282. [3]SHOKROLLAHI A.Raptor codes[J].IEEE Transactions onInformation Theory,2006,52(6):2551-2567. [4]CAO Y,WANG X,LI S.Hierarchical Distribution for Fountain Codes in Geo-Distributed Storage Systems[J].IEEE Transactions on Cloud Computing,2022,10(1):214-226. [5]HUANG C,SIMITCI H,XU Y,et al.Erasure coding in Win-dows Azure Storage [C]//Proceedings of the 2012 USENIX Annual Technical Conference(ATC).USENIX Association,2012:15-26. [6]HUANGC,CHEN M,LI J.Pyramid Codes:Flexible Schemes to Trade Space for Access Efficiency in Reliable Data Storage Systems[J].ACM Transactions on Storage,2013,9(1):1-28. [7]RASHMI K V,SHAH N B.A hitchhiker's guide to fast and efficient data reconstruction in erasure-coded data centers[C]//Proceedings of the ACM SIGCOMM Conference.2014:331-342. [8]SHEN Z,SHU J,LEE P P C.Revisiting Locality in Erasure-Co-ded Storage Systems:An Asymptotic Perspective[C]//Procee-dings of IEEE International Symposium on Information Theory(ISIT).2016. [9]LI J,LI B.Erasure Coding for Cloud Storage Systems:A Survey[J].Tsinghua Science and Technology,2013,18(3):259-272. [10]CHEN Z,ZHANG Y,LEE P P C.Ursa:Hybrid Block-Level and Parity-Level Erasure Coding to Mitigate Long-Tail Latency in Geo-Distributed Storage[C]//Proceedings of the 51st International Conference on Parallel Processing(ICPP).2022:1-11. [11]DIMAKIS A G,GODFREY P B,WU Y,et al.Network Coding for Distributed Storage Systems[J].IEEE Transactions on Information Theory,2010,56(9):4539-4551. [12]XIE X,WU C,GU J,et al.AZ-Code:An Efficient AvailabilityZone Level Erasure Code to Provide High Fault Tolerance in Cloud Storage Systems[C]//Proceedings of the 35th Sympo-sium on Mass Storage Systems and Technologies(MSST).2019. [13]HU S,ZHOU Y,CHEN X,Hierarchical Regenerating Codeswith Optimal Repair Bandwidth for Multi-Rack Storage Systems[J].IEEE Transactions on Communications,2024,72(1):120-133. [14]XU X,ZHANG Y.ACH-Code:Adaptive Cross-tier Hierarchical Erasure Coding for Heterogeneous Data Centers[C]//Procee-dings of IEEE INFOCOM.2024. [15]LIU Y,LI J,LIEW S C.Bandwidth-Adaptive Erasure Coding forHeterogeneous Data Centers[J].IEEE Transactions on Mobile Computing,2023,22(10):5821-5835. [16]ZHOU J,LUI K C S,ZHAO J.Topology-Aware Rateless Co-ding for Dynamic Edge Storage Environments[C]//Proceedings of IEEE INFOCOM.2022:1459-1468. [17]LIANG J,LUI J C S.Geo-Distributed Storage Systems withRateless Codes[C]//Proceedings of IEEE International Confe-rence on Distributed Computing Systems(ICDCS).2023. [18]KARZAND M,LEITH D J,WANG J,et al.On the Decoding Probability of Finite-Length Fountain Codes[J].IEEE Transactions on Information Theory,2020,66(1):190-204. [19]HAYAJNEH K F,SCAGLIONE S V.Analysis of Ripple Size in Luby Transform Codes[J].IEEE Access,2020,8:12345-12356. [20]DIMAKIS A G,KARZAND M.Finite-Length Analysis of Rateless Codes with Application to Distributed Learning[J].IEEE Journal on Selected Areas in Information Theory,2023,3(2):210-222. |
|
||