Stochastic analysis and performance evaluation of wireless schedulers

In the last few years, wireless scheduling algorithms have been proposed by supplementing wireline scheduling algorithms with a wireless adaptation scheme. However, QoS bounds have either been derived for flows that perceive error-free conditions, or a static worst case channel condition. Such an assumption of the channel condition is unrealistic since channel errors are known to be bursty in nature. Hence, these bounds are inadequate to characterize the scheduler’s QoS performance. Our research focuses on performing an extensive analysis of wireless scheduling in order to derive statistical QoS performance bounds under realistic channel conditions. In this paper, we develop stochastic models for various wireless schedulers. Based on these models, we define and evaluate statistical QoS performance metrics in terms of throughput, delay and fairness under various channel conditions and over different time scales. Numerical results indicate that no single scheduler out-performs the others in terms of all the QoS metrics under all channel conditions. The choice of an optimal scheduling mechanism depends on the priority of QoS requirements as well as the channel conditions. Index Terms Statistical QoS, Wireless Scheduling, Fairness.