MassCloud云存储系统构架及可靠性机制  

Architecture and reliability of cloud storage system MassCloud

作　　者：马玮骏[1,2] 吴海佳[1] 刘鹏[1] 

机构地区：[1]解放军理工大学指挥自动化学院,江苏南京210007 [2]解放军理工大学气象学院,江苏南京211101

出　　处：《河海大学学报（自然科学版）》2011年第3期348-354,共7页Journal of Hohai University(Natural Sciences)

基　　金：国家高技术研究发展计划(863计划)(2008AA01A309)

摘　　要：为了解决分布式存储系统的存储容量、可靠性以及效率问题,首先提出了高可靠性海量云存储系统MassCloud的分层体系结构,并在此构架的基础上提出了基于纠删码机制的快速编解码算法——双表法,以及基于纠删码、副本冗余和RAID技术相结合的MassCloud可靠性保证策略,并且进行了测试与分析.结果表明:双表法具有较高的编解码效率,在广域网环境下不会成为性能瓶颈;MassCloud采用4组(8块硬盘)作为每个低功耗存储节点包含的硬盘数,云存储站点层采用3个副本,分布式环境下冗余倍数取2,文件分块数取8,能够保证系统整体可靠性在0.99以上.The hierarchic architecture of the cloud storage system MassCloud with high reliability and mass data storage capability was proposed to solve the current problems of distributed storage system,that is,storage capacity,reliability and efficiency.On such a basis,a double-table method was designed based on the erasure code to gain high encoding/decoding speed in order to suit the distributed storage environment.The reliability of MassCloud was guaranteed by a policy combining the erasure code,duplicate redundancy and RAID.The relevant tests were performed and analyzed.The results show that the double-table method has high encoding/coding efficiency and may not become a bottleneck under the WAN environment.The whole reliability of MassCloud can be ensured to be above 0.99 by using 4 groups of disks（8 pieces of disks） with low cost storage level,3 replicas for cloud storage level,2 times redundancy in distributed environment and 8 pieces of file division.

关 键 词：MassCloud 云存储 纠删码 副本冗余 存储可靠性 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]