Research On Key Technologies Of Load Balancing In Cluster

The number of requests on the Web servers is increasing rapidly with the growth of Internet, so availability and efficient performance are strongly required for the Web servers. Achieving load balancing by request dispatching is one of the basic strategies to improve the performance of the cluster system. The thesis introduces some methods and technologies in load balancing system and investigates how rough set theory can be employed to solve load balancing problem.Rough set theory is a kind of data analysis method to deal with vagueness and uncertainty, a rapid growth of interest can be lately seen in rough set theory and its applications. We expound rough set theory and its method, reconsider the request dispatching in the view of equivalence class. According to different equivalence classes which classify servers by load, we propose the load balancing algorithm RSLB based on rough set. The algorithm realizes the load balancing among servers, prevents the servers to slant and improves the performance of the server cluster system. After analyzing and reviewing the existing request dispatching and load balancing of web server cluster system, it is obvious we should study the advantages and deficiencies of static load balancing and dynamic load balancing respectively. Besides, we set up a load balancing system model, which is based on equivalence class of rough set and integrates the advantages of static and dynamic load balancing technology. The model realizes the request dynamic and efficient dispatching in some extent. At last, we show the performance tests for load balancing algorithm RSLB with RR and LLF. The result shows RSLB has a high efficiency in request dispatching and load balancing, its application in load balancing cluster system is proper and feasible.