军事知识服务平台
纠删码存储系统更新方法综述
A Survey of Update Methods for Erasure-Coded Storage Systems
摘要 大数据时代下,数据量持续增长,存储系统中数据丢失的风险也随之不断提升.传统副本技术虽能保障数据可靠性,却带来极高的存储开销.纠删码技术则能以极低存储开销,实现与副本技术相当的可靠性,有效节省存储资源.但纠删码技术在进行数据更新时消耗大量系统计算和网络资源,任何数据块的更新需同步更新条带内所有校验块,从而维护纠删码条带中数据块和校验块的编码一致性.本文系统梳理了近年来纠删码数据更新方法的研究现状:首先,阐述纠删码的基本概念与更新基础,明确其在当前存储系统中面临的核心科学挑战;其次,从磁盘 I/O 降低、网络流量降低、并行传输优化及系统更新优化四个维度,综述相关研究进展,重点分析各类方法的关键技术特征,对具有代表性的更新方法的性能指标做了对比分析,总结其优势与不足;最后,梳理该领域尚未解决的研究挑战,并展望未来潜在的研究趋势及适配的技术路径.
Abstract:
In the era of big data,data volume continues to grow,and the risk of data loss in stor-age systems is also increasing accordingly.Although traditional replication technology can ensure data reliability,it incurs extremely high storage overhead.Erasure coding technology,by con-trast,can achieve reliability comparable to that of replication technology with extremely low stor-age overhead,effectively saving storage resources.However,erasure coding technology con-sumes substantial system computing and network resources during data updates—any update to a data block requires synchronous updates to all parity blocks within the stripe,thereby maintai-ning the coding consistency between data blocks and parity blocks in the erasure coding stripe.This paper systematically reviews the research status of erasure coding data update methods in re-cent years:first,it elaborates on the basic concepts and update fundamentals of erasure coding,clarifying the core scientific challenges it faces in current storage systems;second,it summarizes the relevant research progress from four dimensions,namely disk I/O reduction,network traffic reduction,parallel transmission optimization,and system update optimization,focusing on ana-lyzing the key technical characteristics of various methods,conducting a comparative analysis of the performance indicators of representative update methods,and summarizing their advantages and disadvantages;finally,it sorts out the unsolved research challenges in this field and looks forward to potential future research trends and adaptive technical paths.
刊 名 计算机学报 ISTICEIPKU
年,卷(期): 2026, 49(9)
分类号: TP393
机标分类号: TP393 P208 TP277
在线出版日期: 2026年10月8日