Design And Implementation Of A Software Tool For IP Network Traffic Analysis And Prediction

With the rapid development of IP network and Internet applications, there are more and morerequirements from the Internet Service Providers(ISPs) to the grainer traffic management. Onone side, the large amount of real traffic data has to be processed and transferred to meaningfulnetwork information, such as the utilization ratio of link. On the other side, the future trend of thetraffic is also required for the further network planning work. To meet with these requirements, ananalysis and prediction tool is designed for IP backbone network traffic in this thesis.This tool include five components, i.e., data preprocessing, node traffic analysis, node trafficprediction, traffic matrix analysis and traffic matrix prediction. The tool takes the traffic matrix da-ta from current network as main analysis data, node traffic which measure by SNMP from networkas auxiliary data. In the module of node traffic analysis, the functions of node traffic modeling,traffic decomposing, traffic comparing, and link utilization calculation are provided. In the mod-ule of node traffic prediction, the prediction models of ARIMA and multiple-variable regress areprovided. In the module of traffic matrix analysis, the observations of node traffic matrix in keyelements and fan out elements are provided. In the module of traffic matrix prediction, the predic-tion results of node traffic can be dispatched to the matrix elements, as well as a visual display incomparing node in-traffic and out-traffic, link load utilization and traffic generated or destined tonodes.The traffic analysis and prediction tool in this thesis has been tested by current networktraffic data providing by a certain ISP in China. The tool can import the original traffic datacollected from real IP network. It also supports the output of graphical reports of the analysisresult. In general, the test results shows that this tool could provide a scientific reference fornetwork maintenance and network planning, and make the positive beneficial effects.