A New Method to Analyze Failure Effects in Pervasive Computing Environment

Module based development take the dominant position in modern pervasive computing system design. Different models and methods have been proposed to estimate effects of module failure on system dependability given the information about system architecture and the interaction mechanism among modules. Most research work put their focuses on stochastic data error which is expected and proved to create unpredictable failures while functional failures are being overlooked. This paper presents a new method based on fault injection to analyze effects of functional module failure on the holistic system. In contrast to failures caused by data error which indicates that transient fault occurs, functional failures are more coincident with engineering practice as a consequence of the complicated interactions among system components. The method has a practical significance especially for pervasive computing systems those are often implemented as embedded systems and have relatively simple architecture diagrams. We demonstrate the analysis process with a case study and its application can extend to real systems.