Impact of Channel Modeling on the Performance of Wireless Scheduling Schemes

Wireless packet scheduling schemes provide the means to achieve service differentiation for wireless users with diverse quality-of-service (QoS) demands. These schemes are often designed to account for the dynamics of the wireless channel, so that a user that is likely to experience deep fade defers his transmission until the channel conditions improve. The vacant transmission slot is then given to another user that is expected to experience a “clean” channel. Channel conditions are location dependent and are made available to the transmitter (the scheduler) after some time delay. Therefore, at the time of packet scheduling, the scheduler has to predict the current channel state at a given receiver, often based on an N -state Markov model. The goal of this paper is to assess the impact of channel predictions on the delay and throughput performance obtained under two popular wireless scheduling schemes. For both schemes and under reasonble channel assumptions, we show that better performance can be achieved by increasing N .