Abstract
M.Ing.
Recent traffic analyses have shown the
existence of long-range dependencies in network traffic,
specifically self-similar long-range dependencies. Due to
the inability of traditional traffic models to capture these
long-range dependencies, new network traffic models
were developed that are able to capture it.
In this paper we compare three self-similar long-range
dependent traffic models, namely the FARIMA model,
the wavelet independent Gaussian model and the
multifractal wavelet model. We present results on their
marginal distributions, their correlation matching to
real traffic and their queuing behaviour. We show that
the multifractal wavelet model is the best of the three
models in all of the test aspects.