The design phase of any real-time system requires balancing the limited computational resources against the functional requirements and the performance of the application. The optimal design solution can be obtained by solving an optimization problem where the system performance is maximized within the schedulability constraints.

In this paper, we provide a procedure that finds the task activation rates maximizing a performance function within the deadline constraints in systems scheduled by fixed priorities. First, we describe the exact feasibility region in the domain of task frequencies. Then, we introduce a procedure that starts by finding an initial solution, and incrementally improves it by using an original branch and bound search, until the global optimum is reached.

Experiments show that our algorithm finds the optimal task periods for practical problems with a remarkable speedup, if compared with existing techniques. When the size of the problem makes the global search intractable, the experiments show that the algorithm can still find a high quality solution in the very early steps.