Predição de tráfego, usando WT e ARIMA, aplicada ao gerenciaciamento adaptativo de largura de banda para interfaces de roteadores

Abstract

This dissertation presents the development of a framework to aid the identification of policies for adaptive management of bandwidth maximum permissible in connection interfaces with LANs on a router. In this work is evident that adaptive management can be performed based on predictions of traffic through WT and ARIMA. We investigated the accuracy of the predicted traffic by comparing prediction methods and observing that the combination of WT with ARIMA is more accurate. Data collection, which originated the time series used to make the prediction was performed using the SNMP protocol and the dynamic reallocation of bandwidth was done through the CBQ mechanism. A contribution of this work is the development of the algorithm SGALB, which can be used as an auxiliary tool for performance management in computer networks.