Architectural Reasoning for Safety- Critical Software Applications
Weihang Wu · 2007
In recent years, there has been substantial move towards architecture-based development for safety-critical software applications. Nevertheless, techniques for architectural design have not been developed to the extent necessary to guarantee the safety of these systems. In particular, current practices often focus upon architectural modelling without making the underlying design deliberation explicit. Although a number of protection mechanisms have been codified in both research and practice, there is little practical guidance mi how to exploit them to address application-specific safety concerns. This thesis defines and demonstrates a coherent and effective approach to eliciting and formulating safety concerns, evaluating and mitigating safety concerns, and arguing about safety within the evolutionary architecting process. The elicitation and formulation of safety concerns are based upon the unification of goals and scenarios from both positive .and negative perspectives. The evaluation of safety concerns is based upon the application of Object-Oriented Bayesian Belief Networks and mitigation space is informed by the evaluation results and the formulation of safety concerns. Safety argumentation of architectures is based upon the definition of a set of patterns of argument. Guidance on safety argument review is also provided. Through addressing safety concerns in the early system development lifecyc1e, there can be an increased level of design confidence in the architectures developed for safety-critical software applications. Evaluation of the approach is conducted through a number of academic and industrial case studies.