Design And Implementaion Of Performance Analysis System For Application Performance Management

Performance analysis is significant for building a highly efficient computing system.However,with the popularity of cloud computing,computing system gets more and more complex,it is even harder to complete performance analysis.Current performance analysis tools mainly adopt behavior simulation method to focus on user experience,monitoring response time,error rate and other parameters,which weakens the analysis section.What's more,most of them only applies to user interactive applications and can not be applied to other fields such as scientific computation.Therefore,the design and development work of novel performance analysis tools is imminent.Based on machine learning algorithms,this paper designs a performance analysis system to dig internal connection between applications and integrate several indicators when locating exception.The system has been proved to solve the problem locating defects and applicable scenario limitation.Firstly,this paper describes the background knowledge and the related technologies,and then summaries the functions and characteristics of current analysis tools.Through analyzing the position of performance analysis system in the whole APM(Application performance management)system,combining with test of RTC(Real Time Communication)application examples,the working scenario and overall demand of analysis system is obtained.Then this paper proposes the integral structure,introduces key processes and are then describes the application interfaces.Based on these,the class diagram is presented.Also,the key issues involved in every module has been designed in details.After that,we evaluate the system through functional testing and validation.Finally,we summarize the whole design work and propose future research directions.