Research And Performance Analysis Of Cluster Server

The paper researches about analyzing and optimizing the performance of server cluster. Firstly, on the base of the technology of cluster server and structure of UEMC system, the improved performance model of UEMC system based on Queuing Theory which combines with network and data duplication, is built. After making brief of models, this paper seeks the expression of average response time with the view of probability, and discusses the parameter influence of model. Secondly a new distributed load-balanced algorithm is proposed to make the server nodes perform the tasks about collecting information and calculating nodes'weight by themselves and response the data to the load-balancer. Therefore the load-balance server's burden is relatively decreased and the cluster's efficiency is increased. Finally, by analyzing UEMC system's user type and request content, a dynamic request scheduling strategy is proposed to change the order of server process request according to the importance of UEMC system's user type and request content. Compared with the other two types of strategy, the results show that the strategy effectively improves the performance of critical requests.