In this paper, we discuss implementation issues of applying a simulation-based approach to asemiconductor fab-level decision making problem. The fab-level decision making problem isformulated as a Markov Decision Process (MDP). We intend to use a simulation-based approach sinceit can break the "curse of dimensionality" and the "curse of modeling" for an MDP with largestate and control spaces. We focus on how to parameterize the state space and the control space.