云计算中基于相似性图的故障自动诊断方法  

Automatic Fault Diagnosis Approach Based on Similarity Graph in Cloud Computing

作　　者：杨艳燕[1] 

机构地区：[1]南阳理工学院计算机与信息工程学院,河南南阳473004

出　　处：《计算机测量与控制》2014年第12期3877-3880,共4页Computer Measurement &Control

基　　金：河南省科技攻关项目(122102210563);河南省科技攻关项目(132102210215)

摘　　要：故障的自动诊断和修复是云计算提供持续服务的关键,为了提高云环境下故障自动诊断的性能,提出了一种包含相似性替代图和检测图的故障诊断框架;首先,根据相似性将系统的运行指标和事件构成替代图,对替代图中的节点进行分组,使得同一组中的节点可以相互替代;其次,根据事件的时间关系将故障表示为事件的序列,通过排名的方法识别出关键事件并以此构成故障模式;最后,提出了一种基于贝叶斯方法的故障诊断算法;实验表明,文章提出的故障诊断方法与著名的fingerprints方法相比,故障诊断的准确性更高,诊断效率更快。Automatic diagnosis and recovery of fault is critical for providing continuous services in Cloud computing.In order to improve the performance of automatic fault diagnosis in Cloud computing,this paper proposes a fault diagnosis framework containing both substitution graph and detection graph.Firstly,based on similarity,we constructed the substitution graph with metrics and events,clustered nodes in the substitution graph,and made sure that nodes in the same group can substitute with each other.Secondly,we constructed the detection graph according to the time stamps of events,and ranked node in the graph to construct fault pattern containing key events.Finally,we proposed a Bayesian based fault diagnosis algorithm.The experiments show that,the proposed approach is higher in accuracy and quicker in efficiency for fault diagnosis than the famous fingerprints algorithm.

关 键 词：云计算 故障诊断 性能监测 故障模式 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]