Contemporary safety standards prescribe processes in which system safety requirements, captured early and expressed in the form of Safety Integrity Levels (SILs), are iteratively allocated to architectural elements. Different SILs reflect different requirements stringencies and consequently different development costs. Therefore, the allocation of safety requirements is not a simple problem of applying an allocation "algebra" as treated by most standards; it is a complex optimisation problem, one of finding a strategy that minimises cost whilst meeting safety requirements. One difficulty is the lack of a commonly agreed heuristic for how costs increase between SILs. In this paper, we define this important problem; then we take the example of an automotive system and using an automated approach show that different cost heuristics lead to different optimal SIL allocations. Without automation it would have been impossible to explore the vast space of allocations and to discuss the subtleties involved in this problem.
Type (Professor's evaluation):
No. of pages: