A dynamic scheduling algorithm and admission strategy for multimedia traffic in broadband wireless network. (Part II: Performance and tight bound)

For Pt.I see ibid., p.1378-83, (2000). We discuss the performance of the QPB algorithm and modification of associated admission criteria when there are large deviations among the average arrival rates or quality of service (QoS) requirements from input traffic streams. By simulation analysis, the asymptotical characteristics of the probability distribution of backlog in a global queue are revealed, and is close to the bound determined by the generalized large deviation technique (GLDT). Furthermore, we present a numerical analysis approach to evaluate the performance and the stochastic tight bound of the QPB algorithm. The approach is particularly effective for scenarios with low packet loss probability. The obtained numerical results can be transformed into performance bounds and stored in a base station database as a criterion of traffic admission. Such a performance bound is normally more tight than the bound from large deviation theory. As an example, we present the numerical results under the assumption that all input traffic sources are independent Poisson processes.