基于对等网络的面向小文件的云存储系统  

Cloud storage system for small file based on P2P

作　　者：张启飞[1] 张尉东[2] 李文娟[1,3] 潘雪增[1] 沈雁[1] 

机构地区：[1]浙江大学计算机科学与技术学院,浙江杭州310027 [2]浙江大学软件学院,浙江宁波315103 [3]杭州师范大学钱江学院,浙江杭州310036

出　　处：《浙江大学学报（工学版）》2013年第1期8-14,93,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60773182);国家"863"高技术研究发展计划资助项目(2008AA010A323)

摘　　要：针对目前主从结构的云存储系统在存储小文件延迟过大的问题,提出基于对等网络(P2P)的分布式云存储系统.通过改进Chord路由算法提高了资源的查询效率,在系统中引入中心路由节点,中心路由节点上存储系统中所有节点的路由和状态信息,使资源查询时间复杂度缩短到O(l),客户端预取中心路由节点数据,从而减少数据操作时的时间开销;系统通过备份的策略来保证数据的可靠性,实现中数据备份数为3;系统实现了文件存储、读取、删除及列目录等基本操作功能.实验结果表明,与Hadoop HDFS文件系统相比,该系统的小文件操作时间减少了一个数量级.A novel distributed cloud storage system based on P2P was proposed aiming at the excessive delay problem in the distributed file system with Master/Slaver structure when manipulating the small files. The resource query time complexity was reduced to O（l） by improving the Chord routing algorithm and adding a central routing node, which stored all nodes~ status and routing information. Furthermore, clients can pre-fetch the data on central routing node, so the time overhead was further reduced. A backup strategy was proposed to ensure the reliability of the data, and the replica number was 3. In this system, a serial of basic functions are implemented, such as write, read, delete, list directory, et. al.. Experimen- tal results show that the time manipulating the small size files reduces an order of magnitude compared with Hadoop HDFS.

关 键 词：对等网络(P2P) 小文件 分布式哈希表(DHT) CHORD路由算法 云存储系统 

分 类 号：TP338.8[自动化与计算机技术—计算机系统结构]