基于Markov链的云平台故障模型与分析  

Fault model and analysis of cloud platform based on Markov chain

作　　者：何利[1] 赵志元[1] 

机构地区：[1]重庆邮电大学计算机学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2013年第5期671-674,685,共5页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：重庆市自然科学基金(CSTC2012jjA40039);重庆市教委科学技术研究项目(KJ120508)~~

摘　　要：云平台中节点数量增加到一定规模后,系统故障率上升,性能不再相应提高。以Hadoop平台为研究对象,建立了基于节点状态Markov链的云平台故障模型,利用该模型对故障导致的云平台扩展瓶颈进行分析,确定了在系统容错负载率限制下的系统最大节点数变化率,在理论上证明了该模型的正确性。When the number of nodes increases to a certain scale, the failure rate of the system rises but the performance of the cloud platform will be no longer improved correspondingly. This paper takes the Hadoop cluster as the research object, establishing a fault model of cloud platform based on the Markov chain model, and analyzing the bottleneck of expansion led by the node failure, determining the change rate of the maximum number of nodes under the limitations of the system fault tolerant load rate. At last the correctness of the model is theoretically proved.

关 键 词：云计算 HADOOP平台 MARKOV链 故障模型 容错率 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]