Self-similarity Research Of Wi-Fi Network Traffic

Network measurements show that network traffic exhibit a nature of self-similarity, which leads many problems in the range of network design, control, analysis and management. Up till now, most network traffic measurements and analysis have been performed on wired networks. With the development of wireless technology, it is of great importance to conduct self-similarity research on the WLAN (Wireless Local Area Network) traffic. As the dominating standard of WLAN, Wi-Fi (Wireless Fidelity) provides users with a convenient, high-speed and cheap broad band access service. It has a bright future. In this dissertation, we studied the analysis, modeling, prediction and control of Wi-Fi traffic.Firstly, we analyzed four series of real Wi-Fi traffic traces collected from inside and outside China. Our analysis results indicate that these traces tend to have a behavior of self-similarity.In this paper, we applied the self-similarity time-series analysis method to the modeling of Wi-Fi traffic, presented the procedure of FARIMA(p,d,q) model building and conducted our research using FARIMA(p,d,q) model. It was demonstrated that FARIMA(p,d,q) model is suitable for describing Wi-Fi traffic, which has the nature of both long-range dependence and short-range dependence.We studied the FARIMA model-based prediction of Wi-Fi traffic on the theory of optimum linear prediction, and compared it with other models, which testified the advantage of FARIMA model.In this paper, we expanded the solution of Intserv over Diffserv to the networks with Wi-Fi access, which can meet the needs of QoS (Quality of Service). In our proposed framework, we studied the prediction-based dynamic bandwidth allocation for Wi-Fi networks using the FARIMA model. Our simulation experiments show that this scheme can significantly decrease the queue length, reduce the requirement of buffer size and the frame loss rate. Finally, we studied the methods for prediction-based admission control.