A Design Approach for Soft Errors Protection in Real-Time Systems

Abstract

This paper proposes the use of metrics to refine system design for soft errors protection in system on chip architectures. Specifically this research shows the use of metrics in design space exploration that highlight where in the structure of the model and at what point in the behaviour, protection is needed against soft errors. As these metrics improve the ability of the system to provide functionality, they are referred to here as reliability metrics. Previous approaches to prevent soft errors focused on recovery after detection. Almost no research has been directed towards preventive measures. But in real-time systems, deadlines are performance requirements that absolutely must be met and a missed deadline constitutes an erroneous action and a possible system failure. This paper focuses on a preventive approach as a solution rather than recovery after detection. The intention of this research is to prevent serious loss of system functionality or system failure though it may not be able to eliminate the impact of soft errors completely.