Stochastic traffic analysis of contemporary internet high-speed links

Abstract

The aim of this paper is to provide a better understanding of the stochastic properties of contemporary Internet traffic. Previous known studies on Internet Traffic are based on samples captured several years ago, where both Internet’s applications and user behavior were quite different from today’s world. In this paper, a stochastic traffic analysis of contemporary Internet traffic seen on high-speed backbone links, where the size of the analyzed samples amounts to the billions of packets, is presented. The probability distribution functions of packet inter-arrival time, number of packets per unit of time, and average packet length were both calculated and analyzed. A wide sense stationarity test for the observed traffic was performed on several time scales. In order to analyze the self-similarity properties of the process, the Hurst index, obtained through the wavelet transform method, was employed. Finally, decimation used a useful auxiliary technique in the process of scale invariance analysis is presented.