A technique to analyze the tolerance to transient overloads of a fault-tolerant real-time system

Abstract

Fault tolerance and real time computing have been traditionally considered as different domains. However, missing a deadline is a fault in a real time system. A real time fault tolerant architecture based on a redundancy executive (RX) is presented. The timing properties of such an executive are predictable. On this basis, a technique for predicting the temporal behaviour of a system, based on fixed priority schedulability analysis, is provided. Moreover this analysis can be applied to real time systems that present bounded transient overloads. In these systems the number of missed deadlines over a given period of time is bounded. The architecture, together with a dual time out scheme, masks both value errors and timing errors. Thus, providing a feasible mechanism for achieving fault tolerance for both the functional aspects and the timing aspects. Its application leads to cost effective systems because the resources do not have to be sized for the worst case and moreover, the response times are sometimes better than in the non fault tolerant equivalent system.