Identifying Critical Architectural Components With Spectral Analysis of Fault Trees

Abstract

We increasingly rely on software-intensive embedded systems. Increasing size and complexity of these hardware/software systems makes it necessary to evaluate reliability at the system architecture level. One aspect of this evaluation is sensitivity analysis, which aims at identifying critical components of the architecture. These are the components of which unreliability contributes the most to the unreliability of the system. In this paper, we propose a novel approach for sensitivity analysis based on spectral analysis of fault trees. We show that measures obtained with our approach are both consistent and complementary with respect to the recognized metrics in the literature.