Due to the fast development of the Internet, the traditional mathematical model has become insufficient to efficiently analyze and forecast the practical network characteristics, the research of die network performance, however, is crucial to Internet development The present work focuses on study and implementation of a system for the network flow forecasting based on the China Education and Research Network (CERNET), the largest nation-wide academic network in China. Measurement of the backbone traffic was conducted, and analyses of the correlation of the long-term measurement results were performed subsequently. Based on the analysis results, it was found that the network traffic was characterized with strong periodicity. Therefore, the network traffic forecast based on the periodical exponential smoothing model is presented and implemented. The error analysis of the forecast results demonstrates the effectiveness of this model. The forecast of network traffic provides as well strong evidences for the inspection of the network abnormities and the optimization of the CERNET network performance.
