Cumulus:一个基于网络编码的分布式文件系统  

Cumulus: A Distributed File System Based on Network Coding

作　　者：陈运海[1] 陈正亮[1] 王钦辉[1,2] 叶保留[1] 

机构地区：[1]南京大学软件新技术国家重点实验室,江苏南京210046 [2]南京陆军指挥学院军事训练与管理系,江苏南京210045

出　　处：《微电子学与计算机》2012年第9期74-78,共5页Microelectronics & Computer

基　　金：国家自然科学基金青年科学基金(60903025);江苏省科技支撑计划(BE2010179)

摘　　要：基于网络编码在并发式数据传输时能够有效地降低网络负载的优越性,应用网络编码优化分布式文件系统的容错问题得到了广泛的关注.为了研究网络编码在分布式文件系统中的应用及其在降低网络负载的有效性,文中基于开源的Hadoop项目设计并实现了面向网络编码的分布式文件系统.该系统通过对存储文件进行切分,然后将这些文件切片进行编码存储,能够以更小的冗余开销获得更好的容错性能,同时,能够利用网络编码的性能提高网络资源利用率.Network coding can reduce network load in the process of concurrent data transmission, which attracts much attention on the application of network coding to optimize the fault-tolerance of distributed file system. Based on Hadoop, an open source project, we design and implement a network coding oriented distributed file system to research the application of network coding in distributed file system and its effectiveness in reducing network load. By segmenting storage files and encoding, storing the file slices, the system is able to achieve a better fault tolerance with less redundancy overhead. Meanwhile, network resource utilization can be improved by taking advantage of network coding performance.

关 键 词：网络编码 分布式文件系统 冗余容错 HADOOP 网络资源利用率 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]