Research And Design On Load Balancing Technology In Virtual Computing Environment

The virtualization technology supplied a good solution for the ulitization of data center andother high-performance computing systems. Due to the uncertainty of the application and thedifference of the server processing capacity, the data center may lose load balancing. The loadbalancing algorithm as the key technology of the virtual applications in multi-machineenvironment, which greatly affects the overall performance of the virtual computing environment.If they lose the load balancing algorithm, there will be unbalancing on single server, while theother server is idle. The single server’s request will lack timely treatment, the quality of it’sservice can not be guaranteed, and in can even cause a server paralyzed.Therefore, it becomes a very critical issue to maintain the load of each server and tomaintain system high efficiency.By analyzing the difference among the current virtual machineload balancing algorithms,this paper designed and implemented the load balancing algorithmpreliminary in Linjiu virtual computing environment, and analyzes the advantages anddisadvantages of the traditional fixed threshold algorithm by experiment, and proposed a loadbalancing method in virtual computing environment.In the method,the strategy divided into twomode,it selects the peak-hours strategy or the idle-times strategy automatically by monitoring theaverage load of the server cluster.The strategy selects the improved fixd threshold algorithmduring the peak-hours which prevents migration clustering by CPU pre-growth technology,andselects the CPU global average algorithm during the idle-hours which is relatively free in acertain degree.The migration can be triggered only when the load difference is too large.Bothalgorithm uses the CPU forecast model which ensures that a small transient spike does nottrigger needless VM migration by predicting load tendency.The specific parameters were selected for the CPU forecast through simulationexperiments,the results shown that the prediction of the CPU prediction model was basicallycorrectly.Simulation experiments shown that the load status of servers was different by usingdifferent load balancing algorithm,the method in this paper could play both their respectiveadvantages and thus maintained the server load at an average level.