中文摘要：

软件自愈分析模型作为自愈技术研究的核心内容之一,为如何实施自愈提供了决策依据。采用随机回报网,建立应用三种不同自愈策略的分析模型,分别刻画不考虑自愈、考虑基于时间的自愈策略以及考虑时间和负载因素的自愈策略情形下系统状态变化过程。数值仿真实验表明：基于时间的自愈策略在选取合理的自愈间隔的前提下在系统稳态可用性方面优于不考虑自愈的情形,自愈是一种提高系统可用性的有效方法;进一步地发现,基于时间和负载的自愈策略在可用性和吞吐率方面均优于基于时间的自愈策略,不同自愈策略对于系统可用性的改善效果显著依赖于系统负载。

英文摘要：

Software rejuvenation analytical models, as one of the cores of rejuvenation technology research, provide a decisionmaking ba sis for implementing the rejuvenation. In this paper we build using stochastic reward nets the analytic models with three different rejuvenation policies, the nonrejuvenation, the timebased software rejuvenation, and the time and loadbased periodic delay rejuvenation, to depict the alternation process of system state in these policies respectively. The results of numerical simulation experiments show： i） With the optimal fixed rejuvenation interval for the VMM and the VMs respectively, the timebased rejuvenation policy outperforms the nonrejuvenation one in terms of steadystate availability. Rejuvenation is an effective way to improve system availability. Ji） Further, the time and load.based periodic delay rejuvenation policy is better than the only timebased one in terms of system steadystate availability and throughput. The system availa bility improvement with different rejuvenation policies obviously depends on system load variation.