Hierarchical Modeling for Reliability and Performance Measures

Abstract

In this paper we model three aspects of fault-tolerant multiprocessor systems and study their influence on both performance and reliability measures in a combined way. These include concurrency, contention and fault-tolerance. Hierarchical modeling allows complex systems to be analyzed easily by splitting the overall model into layers. SHARPE is a powerful tool which allows the use of eight different model types that can be hierarchically combined to obtain a solution for some measure of the whole system. SHARPE also allows general distributions and hence instead of assuming exponentially distributed random variables, we can model more realistic cases. Examples presented here are small for ease of explanation; however, larger problems have also been solved within this framework. This paper thus lays the ground work for modeling existing systems with real data using these techniques.

This work was supported in part by the Air Force Office of Scientific Research under grant AFOSR-84-0132, by the Army Research Office under contract DAAG29-84-0045 and by the James B. Duke Foundation under a Graduate Fellowship to the first author.