Network Traffic Characteristics Analysis And Forecast Research

With the rapid development of network communication technology, more and more application service appears, it brings out very high demands of quality of network service, traffic control and network management. Analysis and forecast of traffic flow is the foundment of network management and performance annlysis. This thesis researches on feature model and forecast of traffic flow based on traffic characteristics.Firstly, the thesis describes the current domestic and international situation of flow research, then analyzes several major characteristics in the traffic flow. The self-similar charateristic of operational traffic is analyzed in detail: including its mathematical description, self-similar judgement, the cause of self-similar characteristic, the impact on network performance. Secondly, the method of network traffic forecast--behavioral modeling is proposed. Thirdly, the thesis discusses the typical forecasting methods and models which are based on short-range dependent or long-range dependent characteristic, and the limitation of these typical traditional business models. Therefore, a new model which is based on wavelet method and time series analysis is proposed. Corresponding mixed model is based on time series components with different characteristics which are proposed in the new model using wavelet decompositon. Finally, the classical FARIMA model and the new mixed model are used to establish model on actual traffic flow data. The thesis comprises the simulation results of these two models. The results show that the mixed model can achieve network flow characteristics decompositon, and more comprehensively describe the characteristics of network business flow, result accurately and effectively.