嵌套的基于时间的软件抗衰策略模型与分析

Modeling and Analysis of Nested Time-based Software Rejuvenation Policy

针对软件在连续运行过程中普遍发生的老化现象,提出了一种嵌套的基于时间的软件抗衰策略,在系统级恢复前先执行多次的应用级恢复。较之周期性系统级恢复,该策略降低了恢复成本;较之周期性应用级恢复,该策略避免了预测失败导致的系统失效。论文采用随机Petri网和有限自动机对策略的实施过程进行建模,计算了策略引发的停机时间和恢复成本,并与传统的软件抗衰策略的结果进行了比较。分析结果表明,嵌套的恢复有利于进一步减少恢复时间,降低恢复成本,提高系统的可靠性。

Recently, the phenomenon of ＂software aging＂, one in which the state of a software system gradually degrades with time and eventually leads to performance degradation or crash/hang failure, has been reported. To counteract software aging, a nested timed-based software rejuvenation policy was put forward. According to this policy, the system executed application-level rejuvenation several times before system-level rejuvenation every time. Comparing with the periodic system-level rejuvenation policy, this policy could reduce the rejuvenation cost further. Comparing with the periodic application-level rejuvenation policy, it could avoid the system outage due to the failure of forecast. Stochastic Petri nets and finite automata were used to model the working process of this policy. The overhead incurred by the policy itself was computed by mathematical tool called Mathematica. The numerical results demonstrate that this policy can reduce downtime and rejuvenation cost, and then the software reliability and availability can be enhanced.