Research And Implementation Of The Classification Analysis System For Applications Running Performance Based On Semi-supervised Learning

With the increasing popularity of large-scale systems and complex applications in recent years,traditional application-based performance analysis and operation based on artificial policies have been difficult to sustain.In the meantime the increasing the amount of data that can be harvested and an increase in the variety of hardware resources make it paramount to analyze the relationships among multiple performance parameters during application execution.Therefore,more automated,more intelligent application performance analysis is the development direction of next-generation application performance management.Based on the semi-supervised learning data mining and machine learning algorithms,this thesis designs a classification analysis system for analyzing applications' performance.The system is suitable for analyzing multi-parameter application data,and supports the behavior pattern classification of hardware resource consumption for target applications of and detection of abnormal behavior patterns.It proved to solve the problem how to describe the application of behavior patterns,and how to identify the application of abnormal behavior problems.It provides the basis for users to further capture anomalies,locate performance bottlenecks and make optimization analysis.This paper first introduces the background knowledge and related technologies,then analyzing the demand decomposition of the classification analysis system for applications' performance in detail.According to the demand analysis,a complete set of typical application performance classification and analysis algorithms is proposed,including adaptive feature extraction algorithm,pattern classification algorithm based on semi-supervised learning,etc.Next,this article describes the application of functional classification analysis of the overall functional structure of the division,data structure design and internal and external structure design,describing the interaction process between different modules and systems in detail,and finally completed the prototype of the system.Through the detailed test cases and test results analysis,in line with the expectations of system requirements and functions,the system is proved to be feasible.