拜占庭容错纠删码分布式存储协议  

Byzantine Fault-Tolerant Erasure Eoded Distributed Storage Protocol

作　　者：蔡鸾佳[1] 

机构地区：[1]同济大学电子信息工程学院,上海201804

出　　处：《计算机系统应用》2012年第2期98-103,共6页Computer Systems & Applications

基　　金：安徽省教育厅自然科学基金(2005KJ004ZD)

摘　　要：研究了从数据密集大部分拜占庭容错分布式存储协议使用复制技术,但是当存储的数据块很大时,复制技术要求大量的存储空间并占用网络带宽,效率低下。采用纠删码技术则将数据块编码为长度相同的分片,然后把各个分片分别存储在对应的存储节点上,这样可以节省存储空间和网络带宽。拜占庭容错纠删码分布式存储需要额外的开销、附加的服务器甚至要版本化存储,才能保证数据的一致性。通过对通常的情况进行优化,采用新颖的机制并引入同态指纹检验码,使得需要最少的服务器、最小的额外计算量和最小的通信回合数目,达到了低开销,并且可以保证时间戳不跳跃。Most byzantine fault-tolerant protocol distributed storage protocol use replication,but if the block is very huge,replication requires network width and many spaces, so its efficiency is low. Employing erasure coding makes the data block coded into fragments of the same length, then store the according fragment on corresponding node. This can save spaces and network width. Byzantine fault-tolerant erasure coded distributed storage usually requires much overhead, extra servers and versioned storage for the consistency of the data fragments. The paper optimizes for the common cases, employs novel mechanism and introduces homomorphic fingerprints. The protocol reaches the real overhead, because it requires minimal servers, least extra computation and minimal communication rounds. Furthermore, it can make timestamp no-skipping

关 键 词：拜占庭容错 纠删码 复制技术 同态指纹 分布式存储 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]