PhD Dissertation Defense of Faisal Iradat

The Public PhD Dissertation Defense of Faisal Iradat is scheduled to take place on Saturday August 17th from 9am to 11am at the FCS Building, Room CC5. Details of the Dissertation are as follows:

By: Syed Muhammad Faisal Iradat
Supervisor: Dr. Sayeed Ghani, Associate Dean, Faculty of Computer Science


This dissertation deals with the accurate backoff process modeling of the IEEE 802.11 Distributed Coordination Function (DCF) and characterization of traffic distribution. The backoff process has a significant impact on the performance of the IEEE 802.11 DCF protocol depending on the offered load conditions on the network; unsaturated or saturated load conditions. In unsaturated load conditions a short backoff process is desirable due to emerging queuing dynamics. While in saturated load conditions where the queue might be full, a long backoff process will help in avoiding collisions. Several attempts have been made to develop an analytical model to correctly represent the behavior of the actual backoff process in unsaturated and saturated load conditions. However what seems to be lacking in existing approaches is the consideration of offered load as the key parameter to modeling the backoff process. Since the offered load has a critical role in the backoff process modeling assumptions, in this thesis a complete mathematical model of the DCF protocol is developed to provide a framework through which the IEEE 802.11 medium access control (MAC) behavior can be evaluated. The motivations for this thesis are enumerated below:

(i) To enhance existing legacy models on the DCF, as upcoming IEEE 802.11 amendments (versions 11ac, 11ad, 11ah, 11af, etc.) are based on it. The current enhanced distributed channel access (EDCA) scheme of the IEEE 802.11-2012 standard [1] is only a technical extension of DCF. It may also be noted that the upcoming IEEE 802.11ac and .11ad standard amendments [2, 3] are also based on the core MAC of IEEE 802.11-2012.

(ii) To revisit assumptions behind existing models in favor of improved accuracy. Despite the vast amount of literature work, issues such as how closely an analytical model captures with the actual DCF backoff process have remained open [4]. One of the key reasons for the lack of accuracy in the models lies in the approximations due to assumptions. Thus, revisiting the assumptions of existing models is still desirable for improved accuracy.

(iii) To study the offered load dependency on service time distribution for selecting the appropriate queuing model. In literature [5, 6] several attempts have been made to study the variation in the mean service time for arbitrary offered loads. However, showing the dependency in the form of patterns could provide a useful tool for selecting the right queuing model and further enhance the accuracy of performance analysis.

(iv) To harmonize and unify disjoint approaches to unsaturated and saturated load analysis. In the literature, the unsaturated [7–13] and saturated load [4, 10, 11, 14–16] performance analysis has been addressed separately and there is a need to have a unified approach.

This thesis seeks to address the issues highlighted above and provides a framework for studying the average end-to-end packet delay for arbitrary loads. Furthermore, the characterization of the service time distribution through the dependency patterns is also discussed.

Keywords: IEEE 802.11, DCF, dependency study, information theory, performance evaluation.