Service Performance Monitoring And Analysis System In Cloud Computing Platform

With the emergence of cloud computing, effectively managing the performance ofweb applications hosting in the cloud computing environment becomes a problem of greatconcern. The current mainstream cloud computing vendors and platforms only providelimited monitoring functions, while the traditional performance management methodssuch as transaction tracking are highly intrusive, so that are not suitable for cloudcomputing environment.Service performance monitoring and analysis system in cloud computing platformimplements a mechanism of performance monitoring and performance analysis for webapplications in a virtualized cloud computing environment. The system achievesfine-grained performance monitoring for virtual machines and applications throughmonitoring agents, provides users a monitoring service covers both system level andapplication level, so that saves the users from deploying and managing a monitoringsystem by themselves. Based on performance monitoring, the system implements amethod of performance analysis for web applications with by use of machine learningtechniques. The basic idea is: to transform the performance analysis problem into aclassification problem by correlating application’s performance state with relevant systemmetrics, then apply feature selection procedure to choose several system metrics which areclosely related with application’s performance state.The functional test shows that the system has implemented the main functionalities ofservice performance monitoring and performance analysis for web applications. Theperformance of induced classifiers is used to measure the validity of performance analysis.The result shows that the classifiers induced with performance metrics selected byperformance analysis service have good classification performance, with classificationaccuracy over80%.