Sizing resources of manufacturing systems is a critical problem. Simulation is a useful solution method but sometimes it is infeasible. Then, simulation based optimization strategies seem attractive. Their utilization requires the identification of a cost to minimize, but the explicit formulation this cost is very difficult in practice. We propose an alternative approach aiming to identify the minimal number of resources of each type while taking into account the predetermined design specifications. The problem is then defined as a constrained multiobjective stochastic optimization problem and it is solved by jointly applying metamodeling, "bootstrap" and multiobjective optimization approaches.