Safety-reliability of distributed embedded system fault tolerant units

In this paper we compare the relative performance of two fault tolerant mechanisms dealing with repairable and non-repairable components that have failed. The relative improvement in the reliability and safety of a system with repairable components is calculated with respect to the corresponding system where the components are not repairable. The fault tolerant systems under study correspond to a flexible arrangement of fault tolerant units (FTU´s) suitable for dependable distributed embedded systems. A simple simulation-based methodology to numerically evaluate dependability functions of a wide variety of fault tolerant units is used. The method is based on simulation of stochastic Petri nets. A set of 15 FTU configurations belonging to five groups is analysed. The methodology allows a quick and accurate evaluation of dependability functions of any distributed embedded system design in terms of the type of FTU (i.e., node or application), replicas per group, replicas per FTU, with or without repair functionality, and shared replicas.