Fractal traffic models for Internet simulation

Abstract

We construct five new fractal traffic models based on the unified framework of fractal point processes (FPPs), and analyze their asymptotic statistical properties. Two of them are on-off types, suitable for characterizing an aggregate fractal traffic source. The other three provide a structure suitable for flow modeling, allowing one to gain a quantitative understanding of how application- or flow-level fractal dynamics (such as user activity, session/flow arrivals, duration, and volume distributions) affect packet-level fractal dynamics. Such a structure, together with analytic results obtained in this work, opens an entirely new horizon that enables us to explain the causes and origins of complex multiple-time-scale behavior of packet-level dynamics in a quantitative manner, i.e., multi-fractal patterns over small time scales and mono-fractal behavior over large time scales. Consequently, they provide network traffic engineering researchers and practitioners with practical and flexible tools for analyzing, devising and performing various traffic engineering studies. All of the FPPs are implemented in OPNET as IP-routable traffic generators.