A fault resilience tool for embedded real-time systems

Abstract

A simulation-based approach to measuring the faultresilience of real-time systems is presented. Simulation is used to favor generality, comparability and make it possible to study the system taking into consideration its overall behavior instead of dealing only with worst-case scenarios. Tasks can be analyzed individually, which may be useful when they have different levels of criticality. The simulation procedure is efficient since only randomly generated parts of the schedule are simulated. We show that results collected from simulation can then be statistically analyzed for different scheduling models so that one can infer the overall fault resilience for the system.