A Model-Driven Approach for Runtime Reliability Analysis

Abstract

Runtime reliability analysis has proven to be a valuable technique to enhance the overall reliability of safety-critical systems. It has the potential to close the dependability gap that has been identified by Laprie. However, existing approaches suffer from either too complex and therefore error-prone input languages or from long execution time due to the state space explosion of the underlying analysis techniques. In this paper, we present an approach for runtime reliability analysis, which handles both problems. It provides a compact metamodel that can be used to describe all necessary information. Moreover, it provides analysis algorithms that can be automatically parameterized by code generation. These algorithms are runtime efficient so that they can be executed even on low-end computers, e.g., safety-critical embedded systems, to adapt the system to changing environmental conditions.