Análise de tendência de tráfego origem destino em redes IP utilizando estatística multivariada

Abstract

Anticipating the trend of the traffic behavior may help the network administration in many of the operational and planning tasks. Traffic through links and nodes is composed by superposition of origin destination (OD) flows, according to a routing matrix, which creates a multivariate structure and high dimension matrices. The analysis at one time of growth trend of OD flows with similar behaviors may help prevent future network congestion. This paper uses multivariate Statistics to analyze and forecast the traffic of OD flows. Previous papers have shown that OD flows can be modeled by using few components, called eigenflows in a principal components analysis. This work proposes a methodology to analyze the OD flows behavior and its growth trend in backbone networks. The experiments performed with actual data from the GEANT network show suitable results.