Priority and dynamic scheduling in a make-to-stock queue with hyperexponential demand

We consider the scheduling problem in a make-to-stock queue with two demand classes that can be differentiated based on their variability. One class experiences Poisson arrivals and other class experiences hyperexponential renewal arrivals. We provide an exact analysis of the case where the demand class with higher variability is given non-preemptive priority. The results are then used to compare the inventory cost performance of three scheduling disciplines, first-come first-serve and priority to either class. We then build on an existing dynamic scheduling heuristic to propose a modification that works well for our system. Extensions of the heuristic to more than two classes and to the case where demand state is known are also discussed.