Reliability and Safety Guarantees in Modern MPSoCs with Real- Time Requirements

Abstract:
Due to continuous technology downscaling modern MPSoCs become more and more susceptible to the occurrence transient faults which may directly lead to internal errors.The usage of appropriate fault tolerance mechanisms is necessary to counteract these errors and thus preventing them from originating a system failure. In this paper we will explore the impact of fault tolerance mechanisms on individual components in MPSoC with real-time requirements. Therefore we initially introduce behavioural models of on-chip components to specify the component’s reaction on errors. Further on we present two different methodologies to derive the reliability from the behavioural models: Monte-Carlo simulation and formal analysis. Finally an exemplary on-chip communication scenario is considered for its reliability under differing operating conditions. As a key novelty we demonstrate how different levels of safety can be reached within a single system-on-chip.