基于负载模型的软件抗衰决策方法

Workload Model Based Method for Decision of Software Rejuvenation

该文提出了一种新的基于度量的抗衰决策方法,把系统的实时负载作为一个重要因素,建立了系统负载的半马尔可夫模型,基于负载模型分别给出了抗衰实施时间和抗衰粒度的决策方法。实验结果表明:与基于时间的抗衰实施时间决策方法比较,基于负载的抗衰决策方法具有更高的预测准确性;同时选择合适的抗衰粒度将有助于提高系统的可用性,降低抗衰成本。

Software rejuvenation is a new technology for fault tolerance to effectively hold back software aging. A decision approach for software rejuvenation is put forward in order to improve system availability. The real-time system workloads are considered as a very important factor, and a semi Markov process is developed to describe the system workloads. Workload model allows us to determine optimal rejuvenation schedules with respect to various system measures. Compared with the method only based on time, the experimental results show that the method based on workload model has the higher accuracy of prediction, and it is also useful to select the proper rejuvenation granularity to improve the system＇s availability and to reduce the rejuvenation cost.