A Formulation and Theory for Delay Guarantees in Wireless Networks
Seminar Room 1, Newton Institute
Delay guarantees have been problematic in networking. The usual focus of theory is only on providing throughput guarantees. Yet, wireless networks will increasingly need to support applications requiring such guarantees, e.g., voice-over-IP, interactive video, and control over networks. We propose a theoretical framework for addressing the problem of delay guarantees in wireless networks that incorporates three key issues - delay, throughput, and channel reliability - in the specification of quality of service.
A somewhat surprising necessary and sufficient condition characterizes when the quality of service requirements of a given set of nodes can be met. It can be checked in nearly linear time, providing a tractable admission control algorithm. Further, there are easily implementable scheduling policies that are feasibility optimal in the sense that they can meet the demands of every feasible set of nodes. The theory can be extended to more general arrival patterns and fading processes, and can also be cast in a utility maximization framework for delay guarantees.
[Joint work with I-Hong Hou and V. Borkar]