Here the paper demonstrates how recent advances in the study of discrete-event stochastic systems provide fruitful results for the modeling, analysis, and design of manufacturing systems. It considers a multistage make-to-stock system where outputs from the final stage are used to satisfy customer demands. The paper addresses the problem of finding the appropriate trade-off between reduced order waiting time and increased process speeds. Using the idea of infinitesimal perturbation analysis, it establishes a simple procedure where sample-path derivatives can be obtained along an arbitrary sample path. Under suitable conditions, it can be demonstrated that these derivative estimators are unbiased and strongly consistent and can be used in a classical stochastic optimization scheme to solve the problem. The role of continuity and convexity on the validity of the estimator is also addressed. Although the focus of this article is not to solve for the optimal solution, a theoretical justification is provided for such a pursuit. The approach is appealing as it is numerically stable, easy to implement, and can be extended to other system performance measures.
