Link characteristics and forecast for Internet users

For a user sharing Internet resources the "Weather Forecast" of the Internet is important. As a user, he has no right to access the status of routers/switches and has not enough information about the overall network. In order to give users useful information and the forecast, this paper proposes a new approach to measure and forecast Internet link status and Local area network (LAN) traffic. In this paper, the design and implementation of a user-oriented, generalized system are proposed for characterizing the links and producing short-term performance forecasts based on historical link characteristics. Pathchar, packet pair are used to measure the latency of hops along a link, and TCPdump is used to collect the traffic of the local LAN. When using these measurement tools in appropriate time scales, we can set up a set of indicators for link status and do prediction for the link characteristics in the future. To improve the accuracy of the estimate of bandwidth, the Extreme value theory (EVT) is deployed. The experimental data show the feasibility of EVT filters. The data show that the EVT results are more close to nominal bandwidth. The FARIMA (p, d, q) (fractional autoregressive integrated moving average) based prediction is also presented to forecast the furure status of links and the local traffic.