Towards Modelling Adaptive Fault Tolerance for Resilient Computing Analysis

Fast evolution of computing systems is a hot topic today that is becoming a real challenge for safety critical embedded systems. For both maintenance and functionalities reasons, over-the-air updates are very attractive for embedded systems manufacturers in many application domains. The challenge here is to maintain dependability properties when facing changes. This is exactly the definition of resilient computing we consider in this work. The implementation of resilient computing relies on fault tolerance design patterns (FTDP) that comply with various types of non-functional assumptions (behavioural assumptions, fault model assumptions, temporal assumptions, resources assumptions, etc.). Despite changes in operation, the efficiency of the fault tolerance mechanisms (instance of a FTDP) depends on the strict compliance with such assumptions. The objective of the paper is to provide a model to simplify the analysis of resilient systems, in particular focusing on adaptive fault tolerant computing. Simple measures are illustrated on evolution scenarii.