Research On SaaS Load Balancing Strategy Based On Petri Net

In today's cloud computing environment,there are more and more tenants,who put forward higher requirements to the performance of the servers and their needs is becoming more and more widely,so,the most bottlenecks will appear in the SaaS side when dealing with tenants requests,with the result that tenants requests cannot be processed effectively and timely and SaaS resources cannot be fully utilized.However,a better load balancing strategy can provide users with higher quality services and improve the utilization of SaaS applications.Thus,it is very important and urgent to study load balancing techniques in SaaS applications.SaaS mainly has the following characteristics distinguishing the traditional software:multi-tenancy,scalability.Based on the position of important feature of SaaS,load balancing strategy was proposed based on the distribution of request and global and local scalability,at the same time,the strategy model was constructed by using stochastic Petri net,which called Stochastic Petri Net model.The load balancing strategy can be analyzed according to this model,furthermore,two execution algorithms are proposed on the basis of SPN model.(1)the first execution strategy:the load balancing scheduling strategy based on multi-tenancy and scalability:LB MS:firstly,tenants choose a server cluster matching with their own needs extremely;when there is no appropriate server cluster,the local or global server cluster should be selected randomly,which can handle their own requests,furthermore;it also compared with RR algorithm,random algorithm and ILCS.(2)the second execution strategy:(LBMS + Min-Connection,LBMS + MC):This strategy is based on the first execution strategy and the minimum connection algorithm,which optimizes the first kind of strategy,when selecting the The optimization of the implementation of the algorithm,in the choice of local or global local or global server cluster,it uses the minimum connection algorithm,thus,it can improve the utilization of the server cluster and reduce the waiting time of tenants.This paper constructs a SPN model based on multi-tenancy and scalability using the stochastic Petri net,at the same time,the model is combined with the above two strategies and simulates by the stochastic Petri net analysis tool.The results show that,LBMS strategy has lower total response time and higher total throughput than RR algorithm,random algorithm and ILCS algorithm;LBMS + MC strategy has higher total throughput and lower total response time comparing with LBMS strategy.