A Framework for Heterogeneous Quality-of-Service Guarantees in Wireless Networks: A Communication-Theoretic Approach

Abstract:

The heterogeneity of the networks in terms of performance measures such as bandwidth, data rate, reliability, quality-of-service (QoS) guarantees, and content such as data, voice, multimedia, have grown tremendously in the last decade. Current approaches in providing different QoS to different users either address the problem at higher network layers (e.g., application layer), or use rudimentary techniques at the physical layer collectively known as "unequal error protection'.

In this talk we ask and try to give a partial answer to the following question: "is there an Information/Communication-theoretic framework for understanding the fundamental limits of heterogenous QoS in a multi-user network?'

A well-known and intuitive result from Information theory says that the users' reliabilities (e.g., quantified by their error exponents) can increase by reducing their transmission rates, i.e., by moving away from the capacity boundary. While this is certainly true, there is a completely new degree of freedom in the multiuser setting that is left unexplored. In particular, we show in this presentation that even if the users' rates are fixed, there is a way to trade reliability between users. Loosely speaking, for a given set of users' data rates, there is some total amount of reliability in the system, which can be allocated among the users. We demonstrate these ideas using as examples the Gaussian broadcast channel and the MIMO fading broadcast channel.

None available