Campuses:

QoS Scheduling of Multiple Users Sharing a Time-Varying Wireless Medium

Friday, August 10, 2001 - 9:00am - 10:00am
Keller 3-180
Alexander Stolyar (Alcatel-Lucent Technologies Bell Laboratories)
We start with the following model motivated primarily by the problem of Quality-of-Service (QoS) scheduling in wireless systems, like CDMA HDR. Multiple data flows (users) share a channel with capacity that varies in time randomly and 'asynchronously' with respect to different users. The problem: How to schedule transmissions so that the maximal number of flows can be served at a desired QoS level.

We identify two classes of scheduling disciplines which are 'throughput optimal', i.e. have the maximum possible stability region. We demonstrate that using these disciplines with the appropriate choice of parameters allows to efficiently control QoS.

Then, to explain nice behavior of throughput optimal disciplines, we consider a heavy traffic regime in a more general model which we call a generalized switch. This model includes as special cases the model of multiuser data scheduling in a multiantenna system, the input-queued cross-bar switch model, and a discrete time version of a parallel server queueing system. We show that, under a non-restrictive in applications 'complete resource pooling' condition, a simple `MaxWeight' discipline minimizes the system workload and induces a 'state space collapse' - in the limit the vector of queue lengths is always proportional to some fixed vector.