This research considers the problem of stochastic control of queueing networks. There currently exist tractable models of stochastic queueing networks that allow multiple customer classes, but they permit only very simple scheduling rules. Thus, extant theory gives little guidance as to how or how much system performance can be improved through intelligent routing and sequencing of jobs. This research will examine a class of network scheduling problems and derive simple heuristics that are asymptotically optimal under conditions of heavy loading. A simulation study will be undertaken to compare whatever heuristics emerge from the theory against others appearing in the literature. In both the theoretical development and the simulation study, special network structures characteristic of semiconductor manufacturing systems will be emphasized.
