Study On Queuing Theory Based Multi-tier Web Application Performance Analysis Model

After years of development, web applications have become one of the mainstream of current forms of network applications. The complexity of web applications rises sharply and the performance requirement of web application also increases. For some high performance requirement system such as e-commerce sites, the poor performance will lead to loss of customers, revenue and other serious damages. In short, the performance of Web applications have become a key factor in the success or failure of the system. Therefore, it has become a hot research topic to analyze the multi-tier Web application performance.However, the existing Web application performance analysis methods can not effectively complete the Web application system performance analysis task because most of these methods didn’t consider the web application performance modeling and analysis problems in virtual server enviroment, despite the fact it is very necessary to study this problem because in the cloud computing environment, virtualization technology can achieve the abstraction of the underlying physical resources, allowing to achieve the dynamic use of resources and flexibility scheduling according to actual needs through dynamic resource allocation and task scheduling and other ways.To solve these problems, this thesis establishes a multi-tier web application analysis model based on queuing theory firstly. Compared to the existing model, our model adds a drop queue to handle concurrency limits at tiers. Based on the model, this thesis proposes the method for the model parameterization, both in the physical server and virtual server environment. To ensure the accuracy of the estimated response time, this thesis proposes an algorithm based on multivariate linear regression to estimate the CPU consumption and I/O consumption, and an algorithm to estimate the request discard rate based on the state space model in the physical server environment, and in the virtual server environment, to ensure the accuracy of multi-tier Web application performance analysis. In the last of the article, we carried out a series of experiments to verify the effectiveness of the proposed multi-tier Web application performance analysis model.