Performability analysis of distributed real-time systems with repetitive task invocation

An algorithm and a methodology for the performability analysis of a class of repairable distributed real-time systems are presented. Distributed real-time workloads generally consist of repetitive concurrent tasks with known cycle and deadline times. The planning cycle of real-time distributed systems, which normally consists of several task invocations, is first identified. A repairable distributed real-time system that can be described by a cyclic Markov reward model between invocation tasks is considered. The performability distribution at the end of the planning cycle is determined by performing repeated convolution of performability densities between task invocations. This convolution operation is efficiently done using the Laguerre coefficients. The algorithm numerically determines both moment and distribution of performability in O(N/sup 3/).<>