A Performance Analysis Of802.11Protocols Based On Stochastic Network Calculus

Wireless networking is one of the exciting developments in the world of networkingtechnology. IEEE802.11protocols have been proposed as the standard protocols forwireless local area networks. In order to provide QoS guarantees, it is necessary tocharacterize the delay and other performance metrics in802.11networks. Networkcalculus is a recently developed approach to take a deep insight into flow problemsencountered in networking, especially on QoS guarantee analysis. Unlike queuing theory,network calculus provides a more general way to describe the arrival and service process.It is convenient to use network calculus to get the delay and backlog bounds of network.Due to the stochastic property of arrival and service process, network calculus hasdeveloped from deterministic to stochastic. The stochastic network calculus is moresuitable for the networks which only provide stochastic guarantees, and it adoptsstochastic arrival curve and stochastic service curve as tools to make the analysis ofnetwork performance more precise, which makes stochastic network calculus has a verygood application prospect.In this thesis, we first introduced the theoretical basis of stochastic network calculus,and then applied stochastic network calculus to the analysis of wireless networkperformance. We derive the stochastic service curve of an802.11wireless node and thev.b.c stochastic arrival curve of fBm model and poisson model. Based on the theory ofstochastic network calculus, we get the stochastic delay and backlog bounds of an802.11wireless node, which make up for the lack of traditional methods such as queuing theoryand deterministic network calculus which can only provide the analysis of average valueand maximum value of delay. Through the results, we can analyze how the parameters ofboth fBm traffic and the setting of802.11protocols impact the delay and backlog of an802.11wireless node. Then, we use matlab to calculate the theoretical stochastic delaybounds and ns-2for simulations. The resulting delay and backlog distribution ownpractical significance to guide the parameters configuration of network applications.