A novel Effective Capacity-based framework for providing statistical QoS guarantees in IEEE 802.11 WLANs

This article proposes a performance model of the IEEE 802.11 MAC layer that employs the notion of Effective Capacity. In particular, the paper establishes that an IEEE 802.11 mobile station can be regarded as a Semi-Markovian bursty server of the On/Off type, with known distributions for the On and Off periods, and subsequently applies known results for Semi-Markovian models to derive the Effective Capacity function of this On/Off server. The general Effective Bandwidth/Capacity theory can then be used for computing buffer overflow probabilities and for employing simple traffic control policies to enforce related QoS guarantees. The policies guarantee a soft bound on the buffer overflow probability and are suitable for real-time traffic control over WLANs. The Effective Capacity model of IEEE 802.11 stations is originally developed by assuming that the other competing stations are saturated. This is a conservative assumption that becomes very accurate in a highly loaded network. Subsequently, the model is adapted to encompass lightly loaded networks as well. In the adapted model, each mobile station directly measures a few model parameters, instead of calculating them on the basis of the saturation assumption, and uses these measurements in the computation of its Effective Capacity function. The theoretical results are checked against simulations, validating the appropriateness of the model.