一种数据网格容灾存储模型及其数据失效模型  

A Disaster-Tolerant Storage Model and a Low Data Failure Model for Data Grid

作　　者：曲明成[1] 吴翔虎[1] 廖明宏[2] 张银[1] 杨孝宗[1] 左德承[1] 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150001 [2]厦门大学软件学院,福建厦门361005

出　　处：《电子学报》2010年第2期315-320,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.60533110)

摘　　要：可靠性较高的数据网格多采用双副本容灾可以保证节点在灾难发生时进行有效恢复,但由于节点数据存储量较大,当节点发生灾难时,从一个节点进行数据恢复速度较慢,导致在数据恢复期备份节点发生失效的概率较大.针对这些问题,本文给出一个容灾存储模型,基于该模型推导出一个数据失效模型,理论证明该失效模型的数据失效概率明显小于双副本容灾方式的数据失效概率,同时在灾难发生时又能达到较快的恢复速度.设计了仿真实验,将模型的数据失效概率与双副本失效概率进行了比较,实验结果与理论推导一致,证明了存储模型和数据失效模型的正确性和有效性.最后给出进一步研究思路.With double copies（DC） for Disaster-Toelerance,the data grid with higher reliability can ensure the effective recovery, reduce the data failure probability（DFP） when the disaster occurs, and can provide greater network bandwidth. But because of the large data storage of the nodes, when the disaster occurs to the nodes, data recovery from one node is a little slow, which leads to the higher failure probability of the backup nodes during the data recovery period,therefore,the data failure probability gets higher.To solve these problems, a Disaster-Tolerant Storage ModeI（DTSM） is put forward, based on this model, a Data Failure Model（DFM） is built. Tneorefical proof shows that the DFP of DFM is lower than that of DC and the recovery speed is faster when the disaster happens. To make a comparison between the data failure probabilities of two models, a simulating experiment is designed. The experimental result is consistent with the theoretical deduction, which proves that DTSM and DFM are correct and effective. Finally a further research idea is proposed.

关 键 词：数据网格容灾 存储模型 数据失效模型 并行数据传输 GRIDFTP 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]