Analyzing Scheduled Maintenance Policies for Repairable Computer Systems

Abstract

For systems that can undergo repair, it is common to use regular maintenance policies as a means to achieve availability requirements. A solution method is developed to analyze various scheduled maintenance policies for repairable computer systems. The analysis is applicable to systems with behavior (exclusive of the scheduled maintenance policy) that can be modeled by a continuous-time Markov process, and thus important characteristics can he included in the model. Furthermore, the assumption of perfect repair, which is unrealistic for most systems, is not used. Both transient and steady-state measures are obtained. The measures considered include expected availability, expected number of unscheduled repairs (repairs that are performed outside of the scheduled maintenance period), and the probability of an unscheduled repair. The solution approach is based on the randomization technique and possesses advantages such as numerical stability and ease of implementation. © 1990 IEEE