An Update Scheme for Network-coding-based Distributed Storage System
LIU Bing-xing1,DENG Yu-hui1,2
1(Department of Computer Science,Jinan University,Guangzhou 510632,China)2(State Key Laboratory of Computer Architecture,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100190,China)
Abstract:Due to the advantages of high reliability and low redundancy in data storage,network coding is introduced into the distributed storage system as one of the storage solution,but the overhead of its update process is high.Some update scheme use logbased storage mechanism just to reduce I/O overhead without considering optimizing the usage of network bandwidth.In this paper,we found that many data blocks are updated for many times after analyzing multiple data sets.We propose an update scheme CUM,using the linear characteristic of network coding and adding a cache region in update node to storage the information of data block and parity block.When the data block update again,the information can be read from the cache region to compute the new parity block,which reduces the amount of bandwidth usage to read data from storage nodes.We also run the cache-based update model on the basis of the NCFS and evaluate the performance of CUM on real data sets.The results of the experiment show that CUM reduces the total bandwidth usage and update time up to 30% and 24.57% respectively and speed up the update process by a factor of up to 1.326 compare to Differential Update Model(DUM).
刘冰星,邓玉辉,. 一种网络编码分布式存储系统中的数据更新策略[J]. 小型微型计算机系统, 2017, 38(3): 645-650.
LIU Bing-xing,DENG Yu-hui,. An Update Scheme for Network-coding-based Distributed Storage System. Journal of Chinese Computer Systems, 2017, 38(3): 645-650.