Research On Dynamic Adaptation Load Balance Algorithm Based On Clustering

In recent years, with the development of information technology, more and more online information and services are provided by Internet, which makes the business flow far beyond the single web server's afford ability. A web server system which can provide a large number of web requesting is needed. Considering the system scalability and cost efficiency, a distributed Web cluster has been appeared instead of a mainframe machine.The distributed structure is used to reduce the load of a single server. It is a serious problem for consideration about how to make scheduling distribution business flow effective and reasonable, which will help servers to solve the problems that some servers are in busy while some are in idle.The related concepts of cluster and load balance are described in this paper; load balance is to increase resource utilization and reduce the response time of the task through balance task. According to the processor nodes binding request task time, the load balance algorithm can be classified into static load balance algorithm and dynamic load balance algorithm. The static load balance algorithm and dynamic load balance algorithm are compared and analyzed according to the system performance, resource utilization,development cost, and so on. Common static load balance algorithm and dynamic load balance algorithm are briefly introduced.The weighted least connection scheduling algorithm is analyzed in this paper,and has been put forward to improve the dynamic feedback strategy based on the weighted least connection scheduling algorithm. Requests are classified into web publishing, web transaction, commerce and multimedia by the algorithm. Cluster analysis of web commerce requests for its complexity and defined hard resource consumption requests. Furthermore load condition of web server is needed to be judged by load balance algorithm, and load weight based on dynamic feedback is designed. Reading performance information from proc file system and formulas is used to calculate load condition and server performance,and then backend server's weight can be calculated. Dynamic feedback strategy is used to feed related information back to load balance server. Probability distribution is used to distribute the client request tasks. The experimental results show that dynamic feedback load balance algorithm based on clustering is effective, which have certain superiority and application value, compared to the traditional one.