EECS CSPL SEMINAR SERIES
WINTER TERM 1996


Kimberly Wasserman

Kimberly Wasserman

Department of Electrical Engineering and Computer Science, University of Michigan



February 13, 1996


Stochastic Flows and Resource Allocation Problems in Communication Networks

Abstract -
The concept of a high-capacity and high-quality wireless communication network supporting full mobility and a wide variety of communication services presents major modeling and design challenges. The realization of such a network requires novel cellular architectures, intelligent mobility management, and efficient resource allocation mechanisms. Among the central issues are adaptive base station assignment, power control, and dynamic channel allocation. Likewise, the realization of a truly integrated high-speed network requires new approaches to the issues of flow control, bandwidth allocation, and buffer management.

In this talk, we consider basic queueing models that capture essential features of resource allocation problems encountered in modern communication networks. We address dynamic resource allocation in the presence of (1) heterogeneous resources and heterogeneous traffic flows, (2) finite capacity buffers, and (3) traffic feedback loops. The principal concern is the long-term asymptotic behavior of each queueing model and the manner in which resource allocation affects stability, throughput, and performance measures such as average delay. Under as minimal a set of statistical assumptions as possible, we characterize simple dynamic policies that achieve maximal throughput and obtain bounds on average queue size.



return to Prior CSPL Seminars homepage