Cost minimization of real-time mission for software systems with rejuvenation

Abstract

This paper considers the optimal rejuvenation policy problem for software systems performing real-time tasks. Due to software aging, the system performance deteriorates with time eventually leading to the system crash, which can be catastrophic for critical applications. To prevent the system crash or minimize its occurrence probability, software rejuvenation has been widely adopted for counteracting the software aging effect but at the cost of extra system overhead and downtime. We derive the optimal state-based rejuvenation policy minimizing the total expected mission cost for software aging systems subject to multiple performance degradation states. The solution encompasses an event transition-based numerical method that assesses the total expected mission cost of a real-time task, covering penalty cost from the mission failure, operation cost of running the mission software and the expected rejuvenation cost. The proposed cost evaluation model can accommodate arbitrary types of state transition time distributions. The suggested model also allows simultaneous evaluations of other system performance metrics including the probability of the successful task completion (i.e., mission reliability) and conditional expected mission completion time given a successful mission task. Examples are presented to demonstrate the proposed methodology and optimizations. Effects of different parameters on the rejuvenation optimization solution are also investigated.