关键词: 软件再生，软件老化，流体随机Petri网，集群

Abstract: Server clustering is an usual method to improve QoS and availability of system. But due to the need for longterm continuous operation to provide services, there exists software aging phenomenon in cluster system, which makes the system service rate is reduced and the failure rate is increased with the time increases. In this paper, the effect of software aging and the effect of failure were all considered, and fluid stochastic Petri net was adopted to model the working pattern of cluster system. Then the analysis methods of probability of each system state were given, and the formulas of system availability and throughout were presented. The availability and throughout were taken as the performance metrics of the cluster system. And some numeric experiments were done. The results show that the method introduced in this paper has the ability to describe system and is extendable. By using this method, the optimal rejuvenation period is gotten.

Key words: Software rejuvenation,Software aging,Fluid stochastic petri net,Cluster