Study On Server Balanced Cluster Based On Stochastic High-level Petri Net Model

The rapid development of network application of the computer,has put forward higher and higher requirement to the server on the network. More and more bottlenecks will appear in the end of the server. Cluster technology is a kind of effective way of realizing the high performance service, but the elasticity and dependability that the cluster load balancing of the server has used and improved the usability that the server uses at double, expand the bandwidth of the server and increase the throughput, strengthen the data handling capacity of the network, raise the flexibility and usability of the network.This paper has introduced the cluster load balancing model of server based on stochastic high level Petri net(SHLPN), this is the foundation that this paper studies balanced cluster. Then put forward on the basis of the cluster load balancing algorithm that is feedback dynamically and asked for task queue dispatcher's algorithm in characteristic based on this comprehensive existing algorithm of model. Assign algorithm consider every real-time load of server and respond the situation on the basis of load balancing request that feedback dynamically, through obtaining the status information of the server periodically , calculate out some right value of each service utensil at present dynamically, adjust and dealing with the proportion asked among the servers constantly, to prevent some servers from still receiving a large amount of request while overloading , thus raise the handling up rate of the whole system and ask for the response time. When carry on fault-tolerant treatment to the balanced system of load based on this algorithm, this paper has backed up redundantly the key equipment namely load balancing device. Two association tactics of algorithm these make server cluster each server entity approach the optimum operation state according to one's own handling capacity constantly. Use the relevant theories rudimentary knowledge of advanced Petri network at random, it provide cluster not balanced performance evaluation index of models this (for instance: Throughput , response time), increase to asking solving the difficulty brought for the scale of solving the model, discussion's every index to model make approximate to ask solving. Finally, how under theenvironment of net does it discuss by this paper reasonable valid distribution, manage and transfer resource in order to realize valid equilibrium load that user ask.This paper has earned on deep analysis to the cluster load balancing technology of the server, put forward method can is it ask for response time and raise systematic utilization ratio of resource to shortening better, worth quite consulting to practical application.