Research Of Request Dispatching And High Availability Of Server Cluster System

With the development of multimedia technology and Intemet, requests to the server become more and more. Therefore, web server systems which have high performance, high availability and high cost-effective, are urgently needed. Technology of server clustering is the effective means of sloving the problem that servers are overloading and providing high-performance server. A server cluster system is made up of a number of computers that are connected through high-speed network. Cluster systems have cost performance, good scalability and fault tolerance and so on, so they have been widely used.The concept, classification, and data sharing methods of the server cluster system are introduced in this paper. The characteristics of every kind server cluster are also stated. After that, load balancing and highly avaliablity of server clusters are deeply researched. Load balancing and Web Qos control are combined at first. The existing heuristic algorithm are improved and optimized, and a new load balancing heuristic algorithm based on Web Qos control is presented. Then A request dspatching model of a server cluster with heterogeneity based on goal programming is developed. The requests are classified according to their nature. The capacities of the servers are the constraints, and minimazing the loss rate of requests is assumed to be a goal. The problem of the requests dispatching of a server cluster is resloved by sloving the goal programming model. Through the simulation analysis, the results verify the validity of the algorithm. They not only allow the servers to perform better, but also meet the requirements of response time of requests. Thus performance of the server cluster is improved.For two-node hot-standby server cluster systems in Active/Standby and Active/Active mode, state management models of high-availability server cluster systems are developed. Then an availability model of two-node hot-standby server cluster system based on Markov process is established. The quantitative relationship between the availability probability of a single server and the hot-standby server cluster system is obtained according to the model. Finally, simulation results verify the validity of the model.