Optimization of partial software rejuvenation policy

Abstract

This paper models a real-time software system performing state-based partial rejuvenations for counteracting performance deterioration effects caused by software aging. The decision on performing each rejuvenation, recovery state and corresponding rejuvenation time depend on both the system degradation level (state) and task operations completed immediately before the rejuvenation action. Full rejuvenations where the system performance is recovered to the peak level appear as a special case of the considered rejuvenation model when the recovery state after performing the rejuvenation is the initial perfect state. We suggest an iterative numerical method based on event transitions for assessing the successful completion probability of a real-time task performed by the considered software system. The proposed method has no limitation on the distribution type of any state sojourn time (or state transition time). We further optimize the state-based partial rejuvenation policy for maximizing the probability of completing a particular real-time task. Impacts of different parameters on the optimization solution are demonstrated through examples, including the discretization parameter used in the suggested numerical algorithm, real-time task deadline, and rejuvenation time parameter.