Traffic Modeling And Characterization For Wireless Local Area Networks

Current trends in wireless local area networks (WLAN) indicate an exigent demand to provide broadband wireless services to growing users of mobile laptop and PDA. To meet this need, it is very important to study the traffic characteristics and traffic model of WLAN, which is the purpose of this thesis.Illumined by the findings of self-similarity in wired networks, and after analysing the real WLAN traffic test results in Stanford University, this thesis discovers, for the first time, that the real traffic in WLAN also exhibits obvious self-similarity.Why does the self-similarity exist in WLAN? One conjecture is that the bottom-level wireless links maybe have self-similarity. To verify whether a wireless link also has self-similar characteristic, some traffic tests are performed over wireless links. In consequence, the traffic between point-to-point wireless devices also has distinct self-similarity, and it is the first reason of the self-similar traffic in WLAN.As we known, the traffics over all of the wireless links are coverged and superposed into the traffic of WLAN. And covergence and superposition is an important condition to form self-similar traffic. So, we think it is another reason of the self-similar traffic in WLAN.Then, this thesis obtains the Hurst coefficients, Information-dimension and Box-dimension of the Standford WLAN traffic traces at different time scales and analyzes the trends and regular rules of the above coefficients results. Accoding to the behaviours of the WLAN users, this thesis elementaryly explains these trends and rules.Moreover, this thesis calculates and plots the multifractal spectrums of the Standford WLAN traffic data. The multifractal spectrums show, for the first time, that the WLAN traffic also has typical multifractal spectrum characteristic.According to the above results, using self-similar theory, this thesis establishes and verifies a close-form approximate formula for PLR (Packet Loss Rate) over a wireless link, whose input variables are Hurst parameter, buffer size, and service rate.At last, this thesis uses 2 fractal-based methods to predict the traffic in WLAN. The accuracy of the methods is verified by comparing with the original data, and the comparing results show that the fractal prediction methods are better than others. Additionally, by overcoming the blemish of one of the old method, this thesis improves the old prediction method and enhances the precision.