In this dissertation, we investigate the issue of Quality of Service (QoS) provisioning for IEEE 802.11 MAC protocols. We propose a multiple class p-persistent version of the IEEE 802.11 protocol to support QoS differentiation in wireless local area networks (WLANs). A theoretical analysis model is proposed to predict the performance of this protocol. Through analytical modeling and simulation, we demonstrate that by assigning appropriate different transmission probabilities to wireless nodes of different classes, it is feasible to provide service differentiation and achieve the targeted throughput ratio among different classes, while at the same time, maximizing the total channel utilization.
We also consider several implementation issues to practically implement the proposed work conforming to the IEEE 802.11 standard. In particular, we derive an approximate solution for the analytical model to reduce the computational complexity and estimate the performance deterioration due to such an approximate solution. We point out that as long as the average packet collision period is much longer than the basic slot duration in the backoff algorithm, (which is true in all the 802.11 family standards such as 802.11, 802.11b, 802.11a and 802.11g, as well as many potential new systems with higher transmission rates,) the loss in channel utilization is limited when the approximate solution is used to determine the protocol parameters. We associate the contention window based backoff scheme in the current 802.11 mac protocol to the proposed p-persistent MAC protocol through a simple relation between the contention window, cw, and the persistent probability, p. We also propose a simple yet effective mechanism to on-line estimate the number of active stations in a QoS basic service set (QBSS) so as to handle the network dynamics and optimize system parameters dynamically.
By extending the analytical model, we study the impact of AIFS values in IEEE 802.11e on rate differentiation between different traffic classes, demonstrate that the per flow throughput ratio between different traffic classes is a function of both the persistent probability (or the contention window size) and AIFS values. Under the asymptotic condition, we derive a closed-form relation. Through analytical modeling and simulation, we evaluate the effect of tuning AIFS values to provide throughput differentiation. We recommend to only adjust the persistent probability p or equivalently the contention window size cw to achieve the desired rate differentiation, while keeping the AIFS values of all traffic classes to the same value (e.g., at DIFS). This keeps the dimension of design freedom small and the parameter fine-tuning process tractable.