Research Of Load Balancing For Network Services Based On Semanteme

CNNIC reported that about 20 percent of contents stored in Web servers are accessed by more than 80 percent clients. For improving information server system performance, cluster technology has been brought forward and scheduling strategy for load banancing in cluster is needed to meet the increasing collecting access. The thesis is put forward for the purpose of it. At present, load balancing technology and realized methods based on network layer of network reference model mostly are far beyond intelligence. They can not work when some information is accessed at the same time, and what' s more, have nothing to do with QoS in certain situation. A load balancing model based on application layer is given in the thesis. Through analyzing datagram and distinguishing its application layer semanteme from Virtual KB records automatically, system performance. will be improved by increasing Cache' s hitting the target rate. For preventing dispatcher from becoming access bottleneck, mixed equilibrium strategy is introduced in this model. Strategy of driving demand for state changing is used to deal with load~nonbalancing among nodes due to load-accumulating effect. Sender-table and receiver-table are also presented to implement distributed load migration. The distribution of loads is done intelligently and network performance & QoS are improved in this model. Meanwhile, it will promote the combination and development of network QoS and Web QoS.