Time and Prediction based Software Rejuvenation Policy

Operational software systems often experience an “aging” phenomenon, characterized by progressive performance degradation and a sudden hang/crash failure. Software rejuvenation is a proactive fault-tolerance strategy aimed to prevent unexpected outages due to aging. Existing rejuvenation approaches can be largely classified into two types: Purely Time based software rejuvenation policy (PTSRP) and Purely Prediction based Software rejuvenation policy (PPSRP). In this paper, combining the merits of these two policies, a new rejuvenation policy, Time and Prediction based Software Rejuvenation Policy (TPSRP), is proposed. In this policy, time based rejuvenation policy is performed from the system´s start or restart, during which prediction based policy is also employed. To evaluate the effectiveness of this new policy, system availability and Downtime cost is adopted and a stochastic reward net model is built. Numerical results show that under the same conditions, TPSRP can achieve higher availability and lower downtime cost than both the PTSRP and PPSRP.