Dependability Modeling of Real - Time Systems Using Stochastic Reward Nets

Dependability modeling plays a major role in the design , validation and maintenance of real-time computing systems. Typical models provide measures such as mean time to failure, reliability and safety as functions of the component failure rates and fault/error coverage probabilities. In this paper we propose a modeling technique that allows the coverage to be dependent upon the local (i.e. embedded at task level) and global (i.e. available at system level) fault/error detection and recovery mechanisms. This approach also ensures important savings in terms of the simulation time required for deriving the coverage probabilities. Stochastic reward nets are employed as a unique dependability modeling framework. For illustrating the usefulness of this technique we analyze dependability of a railroad control computer.