Improvement And Implementation Of Research On Dynamic Load Balancing Algorithm Based On Web Server Cluster

When a single web server can not meet user’s requirements of accessing, it is a important way to construct high-performance server with a cluster system assembled with multiple servers to manage the external request together. However, with the exponential growth of internet accesses,web cluster server faces severe accessing pressures,at this time,the load balancing plays an important role to improve the web server cluster’s overall performance. Load balancing problem has always been the key issue of the web server cluster, and its scheduling policy shows direct impact on the overall performance of the web server cluster. There are following problems in existing load balancing algorithms,the reception capabilities is limited during the request receiving phase,and the high load calculation overhead,the low cache utilization,the serious load oscillation occur during the request forwarding phase.In response to these problems, this paper presents an improved dynamic load balancing algorithm, the main work is as follows:First,according to the limited reception capabilities in balancing strategies,this paper investigates the current task scheduling model intensively,and a thread-pool model receiving requests is realized basing on the epoll model.The ability of receiving task requests is improved,for the time and space consumption is reduced, deriving from creation and destruction of processes and threads.Second,about the high load calculation overhead,the low cache utilization,the serious load oscillation during the requests forwarding phase in balancing strategies,an improved dynamic load balancing algorithm is proposed in this paper.The algorithm introduces the thought of request classification to the entire process of request forwarding,which includes the collection and calculation of load,the determination of load state,the selection of the target server,and so on,improving the cache hit rate.During the load collection,this paper considers the node capacity and real-time load utilization,with periodic feedback and distributed computing,whick not only reduces the load calculation overhead,but also ensures the real-time and effectiveness of the load.In the aspect of determination of server state,we combined RED ideology with dual-threshold method,reducing the load oscillation.During the selection of the target server,the introduction of the locality principle and the probabilistic forwarding are used,so the requests are forwarding to the target server reasonably,and the manager can dynamically adjusted the server group based on actual users’needs.Thirdly, this paper achieves a dynamic load balancing system in the web cluster system constructed by linux server. Through comparative analysis and experimental verification with the typical load-balancing algorithm,the dynamic load balancing system not only improves the overall throughput of the web server cluster, but also improves the load balancing degree.