Modeling of high-speed wireline and wireless network traffic

Wireline data traffic, as opposed to traditional voice traffic, has been shown to be self-similar and impulsive. Both characteristics can degrade network performance by causing large delays, packet dropping and by requiring large buffers. Traditional teletraffic models are short-range dependent, thus cannot capture the true traffic characteristics, nor adequately predict network performance. Realistic models that capture these characteristics have been subject of extensive research.; With the increasing demand for wireless Internet access and the fast evolution of wireless techniques, various high-speed multimedia services (e.g. video, teleconferencing, Internet access, voice communications, interactive applications) will soon be provided via wireless networks. To date there have not seen any convincing studies to determine whether wireline traffic characteristics can propagate to the wireless networks.; In this dissertation, we first develop a constructive model for wireline traffic, namely the Rate-Limited Extended Alternating Fractal Renewal Process (rate-limited EAFRP). This model captures the distinctive two-slope behavior of the log-log survival function of real multi-user traffic. As the number of users increases the model results traffic that is Gaussian, which is consistent with real network measurements. We also provide a queuing analysis of the proposed model, which achieves a closer approximation to the observed reality than previous models.; Second, assuming seamless connectivity between wireline and wireless networks, we study analytically the propagation of self-similar traffic through a gateway that interconnects wired and wireless networks. Our analysis concludes that under certain conditions the self-similarity can be preserved through the gateway, while in cases such as non real-time variable-bit-rate (nrt-VBR) multimedia traffic, the self-similarity maybe disappear. We derive bounds of packet processing time in the gateway when fed with self-similar traffic.