Cluster Server Performance Analysis And Research

It is a networking-centered information era and people pursue faster access, more reliable and more functional information services. The increasing popularity of the World Wide Web leads to an exponential growth of Internet usage. On the other hand, the excessive increase of traffic on the World Wide Web makes it hard to predict the increase in client request to popular Web sites. Site managers have to face the problem to improve server capacity. Therefore, the Internet research community encounters a number of challenges of evaluating server performance and of providing a series of high-performance technology for designing Web server to enhance Web QOS. The performance evaluation of cluster Web server addressed in this paper is a step towards ameliorating Web QOS.The main topics of this paper are cluster server architecture research, web server workload characteristic analysis, server capacity analysis and request priority scheduling. Firstly, after studying current cluster web server architecture, we present a novel model of cluster server performance analysis桟WSS (Cluster Web Server Structure) model, which is based on open queuing network model. Some server performance related information could be obtained by applying the model to simulation analysis, for example the relationships among server performance, server architecture, implementation policy and server workload. Secondly, we analysis in a deep-going way locality and whole characteristic, of server load, which is a base of server software realizable technology. So there are important practical and theoretical significance to accurately evaluate the variation of server load. Thirdly, this paper introduced the concepts concerning server capacity and discussed capacity planning & testing. A prediction model of adequate capacity was proposed by logistic rule, which is a cursory quantitative analysis model. We use this model to predict server size and forecast workload. Finally, by analyzing relationship among response time, client request arrival rate, thread number and scheduling policy, Dynamic Priority Scheduling Policy is proposed to change the order of server process request. Compared with First Come First Service (FCFS), Short Request First (SRF) and Dynamic Earliest Deadline First (DEDF) policy. The results show that the Dynamic Priority Scheduling Policy can enhance the Web QOS by increasing just a little response time.