Researches On Fractal-based Network Traffic Modeling And Queuing Performance

The real time steaming media services are always expected widespread developing in the new generation network, therefore, the "High Performance Broadband Information Networks" Demonstration Project (3Tnet) was deploied. The 3Tnet ensures the one-off access and convergence of steaming media services through one stage packet switching in the access layer by the integrative network access equipment. In the core network, the 3Tnet is constructed by the high performance network IP packet switch equipment supporting traffic burstiness. Network traffic can fully reflect the interaction between the network services and the network. Thus, quantitative investigations and analyses of the aggregated network traffic characteristics of the core network are essential and imperative. Moreover, the problem of the complexity of network traffic and its impact on network performance has attracted many researchers.In this thesis, the requirements of 3Tnet for aggregated traffic in core networks characteristics are gerneralized. And, this paper explores a mathimatical model for aggregated network traffic, analyzes traffic characteristics and their impacts on the queuing performance of the network equipment based on the fractal theory. Typical network traffic models and their parameters estimating methods within the framework of self-similar theory are comprehensively evaluated. With the fundamental block and some illuminations provided by the above results, a joint multifractal (JMF) traffic model is proposed. The JMF model can effectively capture the burstiness of aggregated network traffic both in small time-scales and in large time-scales, it is validated by the simulation of the actual network traffic data. The statistical characteristics and the effectiveness of the JMF model are analyzed by the theoretical methods. Some detailed features of network traffic are also revealed through the methods of the actual network traffic data fitting. A method to describe the aggregating process of traffic, called multi-scale burstiness constraint nonstationary (MBCN), is proposed, which takes into account the aggregating mode of real traffic, and generalizes the convergence mode of the 3Tnet. The MBCN method can describe the convergence mode of real aggregated network traffic mathematically. The MBCN method also validates the effectiveness of the characteristic functions provided by the JMF model. Finally, an analytic calculating framework of overflow probabilities for single queue driven by a general multifractal input process is proposed based on fractal characteristics of traffic.The main work and contributions in this paper are outlined as follows: Typical network traffic models in the framework of the self-similar theory are comprehensively evaluated. Besides introducing academic foundation of these models, the paper analyses their backgrounds and application conditions respectively, and compares their inherent advantages, disadvantages and perspectives in detail. The conclusion is drawn as follows: These models can in a sense fit the long-range dependence and the burstiness of traffic in long terms. However, none of them can depict the complexity of traffic in multi-scales. Especially, all these...