Numerical Evaluation of Performability and Job Completion Time in RepairableFault-Tolerant Systems

摘要

Fault-tolerant computer systems change their level of performance (e.g., mode ofoperation or service rate) in response to different events such as failure, degradation or repair. We present a unified model for the analysis of job (task) completion time and the accumulated service (Reward) until a given time (also known as performability). In prior work, the evaluation of the distribution of performability was restricted to nonrepairable systems (represented by acyclic Markov chains). In this paper, we describe an algorithm for the numerical evaluation of the distributions of performability or job completion time, in repairable fault-tolerant systems (represented by cyclic Markov chains). We demonstrate the feasibility of our techniques by means of numerical examples. (Author) (KR)