Abstract
This paper proposes a novel scheme for monitoring and analyzing the performance of Internet applications based on statistical analysis. In our study, an Internet application is considered as a "black box" as the internal architecture and the operation/data flow of most of the applications is often unknown to enterprises or internet service providers. Our scheme discovers the underlying relationship among performance variables in a complex IT environment and builds a correlation model and regression model which can be used to effectively support the proactive performance management and optimization, as well as the root cause analysis to find faults and/or performance bottlenecks for an application. The validity of the models is verified by applying them in a real IT network of China Mobile (Group) Guangdong Co. Ltd., the largest subsidiary of China Mobile.