The objective of this paper is to introduce dependability as an optimization criterion in the system-level design process of embedded systems. Given the pervasiveness of embedded systems, especially in the area of highly dependable and safety-critical systems, it is imperative to directly consider dependability in the system level design process. This naturally leads to a multi-objective optimization problem, as cost and time have to be considered too. This paper proposes a genetic algorithm to solve this multi-objective optimization problem and to determine a set of Pareto optimal design alternatives in a single optimization run. Based on these alternatives, the designer can choose his best solution, finding the desired tradeoff between cost, schedulability, and dependability.