Fall 2009 Seminar Series

DATE: December 10
LOCATION: E51-325
TIME: 4:15pm
Reception immediately following in the ORC ConferenceRoom, E40-106

SPEAKER:
Wuqin Lin

TITLE
beta-Maximum Pressure Policies in Stochastic Processing Networks: Heavy Traffic Analysis

ABSTRACT
Stochastic processing networks provide a powerful abstraction for a wide range of real-world systems. This work is about finding good operational policies for such networks. I will begin the talk by introducing several stochastic processing network examples in manufacturing, service and information systems. Then I will describe a family of operational policies called beta-maximum pressure policies and show that they are throughput optimal (rate stable). The focus of this talk is the performance of the beta-maximum pressure policies in heavy traffic. I will establish a (multiplicative) state space collapse result that the queue length process of the networks can be lifted from a lower-dimensional workload process in diffusion limit. I will also show that for networks that have a single bottleneck resource, a beta-maximum pressure policy is asymptotically optimal for both the workload process and a certain holding cost structure.