A stochastic reward net model for dependability analysis of real-time computing systems

Dependability assessment plays an important role in the design and validation of fault-tolerant real-lime computer systems. Dependability models provide measures such as reliability, safety and mean time to failure as functions of the component failure rates and fault/error coverage probabilities. In this paper we present a decomposition technique that accounts for both the hardware and software architectural characteristics of the modelled systems. Stochastic reward nets are employed as a unique modeling framework. Dependability of a railroad control computer, which relies an software techniques for fault/error handling, is analysed as an application example.<>