The Study Of The Load Balancing Based On Web Server Cluster

The exponential growth of web-based application, coupled with fast growth of Internet, has created the need for more and faster web servers to serve the need of internet users. Web server cluster is a promising solution to achieve high server performance with low-cost computer system.The past several years have seen the emergence of some commercial solutions.But web cluster is only in its early stage in china.The paper classify web server cluster architecture by focusing on the entity which distributes the incoming requests among the servers into three main classes:DNS-based(the earliest research),server-based,transparent clustering(dispatcher-based). The dispatcher of transparent clusering receives service requests from clients and selects a server from the server pool by some scheduling policy to process the request.Based on OSI protocol stack ,L4 clustering (operating at layer four)and L7 clustering (operating at layer seven).Load balancing is the important technology of web cluster. L4 clustering can't get the contents of the client requests . L7 clustering requires the client requests go all the way up to application layer. The extra distance provides us important additional information about the request, i.e. the contents of the client requests at this stage. This can be used to load balancing to get "true balancing ".It is main research areas of current scheduling policy for web server cluster.The followings are the contents of study: (1)The study of web server cluster system architecture.The author analyzed the main technique and the characteristic of web cluster,and classify web server cluster architecture to study.The paper has compared the commercial products in the world.(2)The study of load balancing based on web server cluster.The paper outlined the main characteristics,and analyzed the famous scheduling policies for web server cluster in the world. The paper give one improved algorithm on LARD, it can better balance the load, and provided a trace-driven simulation.(3)The paper provided a content-based request distribution,classify web service into two main categories:web publishing(providing static information and lightly dynamic request)and web transaction(providing dynamic content generated from database queries and secure information). Based on the analysis of server performance ,critical states of web cluster and client requests types, the paper proposed a new scheduling policy considering heterogeneity server performance to each categories web service.The new scheduling policy can balancing the load and achieve good web server cluster performance.(4)At last,the paper focuses on the application of web server cluster technique .Based on the analyse of enterprise E-commerce,studied the web server cluster system architecture,and provided a new scheduling policy which supply QOS service.It has reference sense for the future application study.