Performance analysis of multiple classes of traffic in wi-fi networks: theoretical and experimental study

This dissertation develops analytical models for performance evaluation of Wi-Fi networks (IEEE 802.11e/g wireless LANs) and designs admission control strategy that guarantees the quality of service (QoS) requirements of the accepted traffic. Guaranteeing QoS is a major challenge, particularly for real-time multimedia traffic such as voice and video. The QoS measures of interest are throughput, delay, collision probability, and packet loss probability. Our analytical models accurately capture the complexity of multiple classes of traffic in Wi-Fi networks; we also attain several valuable insights about the performance of IEEE 802.11 WLANs. Our study is based on the infrastructure wireless local area networks (WLANs) setting, in which mobile stations (STAs) access a high speed LAN (e.g., Eethernet) via an access point (AP). After deriving analytical models for interactive voice calls and video-conferencing traffic, we extend this model to include transport control protocol (TCP) for file downloads. Our model includes downlink video streaming traffic. We estimate the maximum number of voice calls that can be supported with and without TCP, with and without video; and additionally the maximum video streaming rate when combined with voice over IP (VoIP) calls and TCP controlled file transfers. We also analyze the effect of arbitrary queue size on the voice capacity, video-conferencing capacity and throughput, TCP and video streaming throughput; and determine bandwidth utilization for variable packet arrival rates. Our proposed model is generic in the sense that it provides a high level system view approach, and some of the existing approaches for Wi-Fi performance analysis become special cases of our model for specific queue sizes. The underlying method of our analytic model is to identify an embedded discrete-time Markov chain at random times. The Markov chain consists of the number of nonempty queues or mobile STAs, the number of packets is waiting therein and the variable to keep track of which traffic classes are allowed to attempt in a channel slot. Here, the channel slot is defined as the interval between two successive channel activities, which are any sort of action carried out by a mobile station at any time. The stationary probabilities of all the Markov states are derived by solving a set of linear equations, thus obtaining the desired performance measures of IEEE 802.11 WLANs. Finally, we use the Qualnet simulator to simulate the Enhanced Distributed Channel Access (EDCA) model according to the IEEE 802.11 specifications. We also conduct an experimental study of the EDCA mechanism of the IEEE 802.11 standard with a real-environment testbed consisting of one access point and 11 laptops to further validate the analytical model and simulation results.