Recovery block fault tolerance considering correlated failures

Abstract

Software fault tolerance methods have been proposed to achieve high reliability. However, these methods suffer from the possibility of correlated failures, where the failure of multiple components leads to system failure. Furthermore, previous methods to assess the impact of correlated failure on software fault tolerance require extensive testing data and are therefore less suitable to conduct reliability analysis in the early design stages. This paper presents an approach to quantify the reliability of the recovery block method to software fault tolerance in terms of the reliability of the components and the correlations between the failures of these components. The approach is demonstrated through a series of examples. Our results indicate that the approach can quantify the negative impact of correlation on the recovery block. Thus, the approach can be used to identify correlations that may impede system reliability.