A Multi Variable Optimization Approach for the Design of Integrated Dependable Real-Time Embedded Systems

Abstract. Embedded systems increasingly encompass both dependability and responsiveness requirements. While sophisticated techniques exist, on a discrete basis, for both fault-tolerance (FT) and real-time (RT), the composite considerations for FT+RT are still evolving. Obviously the different objectives needed for FT and RT make composite optimization hard. In this paper, the proposed Multi Variable Optimization (MVO) process develops integrated FT+RT considerations. We introduce dependability as an initial optimization criteria by confining error propagation probability, i.e., limiting the interactions. Subsequently, quantification of interactions together with RT optimization by minimizing scheduling length is developed. A simulated annealing approach is utilized to find optimized solutions. We provide experimental results for our approach, showing significant design improvements over contemporary analytical initial feasibility solutions.