A New Dynamic Loading Balance Algorithm Based On LVS Cluster

With the popularity of Internet, the population of the users has continually increasing. The hits of some hot webs over a hundred million times one day. Only using one server to afford service for all users is definitely impossible, they can adopt load balancing technology as flux distribution in different servers of offering the same service.The background of this dissertation is to find a solution to resolve the overload of proxy server and www server. In this dissertation, it first introduce the background of load balance. Then, a detailed explanation was delivered on Linux Virtual Server's architecture, working mode, characters, algorithm analysis and core theories. Presently the big problem with load balancing technology in LVS cluster system is: In the process of making load balancing, server nodes can't dynamically and effectively give load feedback to the load balancer, thus unable to play its role very well in balancing. To solve this problem, the paper mainly did some research work in the following respects; generalize and contrast the present web server cluster technology; Conduct a profound study of the theory and implementation mechanism of LVS cluster, explain the structure of nuclear software IPVS in LVS and the implementation process of each module, and analyze its usual load balancing algorithm as well as pointing out its weakness; present a new load balancing algorithm with dynamic load feedback mechanism pointing to the weakness of current algorithms used in LVS cluster; Combine task dispatching based on connections and requirement for timely reflecting server load dynamic change by adopting dynamic loadable module technology, thus implementing dynamic load balancing in the kernel space of operating system.; Construct a LVS cluster system, and test the load response in real experimental environment, which testifies the algorithm is effect.Research result demonstrates that LVS can effectively schedule the loading of servers when heavy traffic. It not only shortens the time of response but also guarantees the service quality.