Research On Load Balancing Of Enterprise Server Cluster Systems

With the increasing scale of enterprises, production lines and departments getting increased, enterprise information sharing and exchanges have become more difficult. The failure of information flow make departments become "information islands", this weakened the ability of departments and the collaboration between them directly, eventually lead to low production efficiency. In order to avoid this phenomenon, enterprise information is imperative. However, the information flow it brings cause substantial increase in the load, which bring enormous pressure on the server. The most effective means to resolve the problem at present is server clustering technology, and load balancing is the core of the realization of server cluster system. The paper get a steel company's information projects as back ground, make a deep discussion on load balancing server cluster system of enterprise.Research on load balancing at home and abroad are mostly confined to the study of algorithms and the strategy and proposed improvement, which limited the improvement of load balancing to a large extent. What's more, the corporate network shows characteristics of multi-level and mandate constraints, algorithms and strategies alone become not so useful. Therefore, a load balancing scheduling system is built to solve the multi-layers of enterprise network. As to the collaboration and competition between enterprises'levels, the system built introduce multi-Agent to get it done. Firstly, a load balancing scheduling system of quadruple is constructed in the paper, which get an improvement on efficiency by combining the simple model of load balancing and the characteristics of enterprise networks; Secondly, in order to enhance the cooperation of the enterprise server clustering, multi-Agent theory is studied, and use blackboard theory to load balancing scheduling system as the information center; Thirdly, the core of the system——assessment and decision-making Agent is designed on the basis of Least Connections algorithm and Weighted Round-Robin algorithm, which has smaller load variance.Finally, a simulation system is designed to verify the load balancing system designed get the load balancing performance; through a further comparison test result with LC and WRR, a conclusion that the algorithm designed has smaller standard deviation, which means can balance cluster load more efficiently is verified..