Pricing and resource allocation in communication networks and supply chains

We consider pricing and resource allocation decisions in stochastic networks that provide Quality of Service ( QoS) guarantees. Such networks model several networked systems, including communication networks that support real-time services and supply chains that emphasize customer satisfaction. We focus into two instances of these problems: (i) revenue or welfare maximization in QoS-capable communication networks, and (ii) inventory control in supply chains subject to given QoS requirements.; Regarding problem (i), we study pricing in communication networks with fixed routing that offer multiple classes of service. Prices for these services can depend on congestion conditions and affect user's demand. Our main result is that static pricing is asymptotically optimal in a regime of many, relatively small, users for both objectives of revenue and welfare maximization. In particular, the performance of an optimal (dynamic) pricing strategy is closely matched by a static pricing policy which is independent of congestion conditions. Our analysis reveals the structure of the asymptotically optimal static prices. Using this structure, and employing a simulation-based approach, we can efficiently compute an effective policy for large networks, even away from the limiting regime. For the simpler case of a single-node problem, we also develop an approximate dynamic programming approach to compute near-optimal policies in large systems.; We further extend our setting by considering demand functions that allow one service class to serve as a substitute of another. For such networks, under certain conditions, we also show that static pricing is asymptotically optimal in the same regime of many small users.; Regarding problem (ii), we study QoS-capable supply chains consisting of a tandem of production facilities (stages). Unsatisfied external demand is backlogged. We quantify QoS by the stockout probabilities at various stages. We propose production policies in two separate cases: when each stage (a) has only local inventory information, and (b) has knowledge of the total downstream inventory. In case (a) the proposed policy guarantees service level requirements. In case (b) the proposed policy minimizes expected inventory costs subject to QoS constraints. In both cases policy parameters are obtained analytically, based on large deviations asymptotics, which leads to drastic computational savings compared to simulation. Our model can accommodate autocorrelated demand and production processes, both critical features of modern manufacturing systems. We demonstrate that detailed distributional information on demand and production processes, which is incorporated into large deviations asymptotics, is critical in inventory control decisions.