Wavelet-based queuing analysis of Gaussian and non-Gaussian long-range-dependent network traffic
Ribeiro, Vinay Joseph
In this thesis, we develop a simple and powerful multiscale model for the synthesis of nonGaussian, long-range-dependent (LRD) network traffic. The wavelet transform effectively doecorrelates LRD signals and hence is well-suited to model such data. However, wavelet-based models have generally been used for meodeling Gaussian data which can be unrealistic for traffic. Using a multiplicative superstructure atop the Haar wavelet transform, we exploit the decorrelating properties of wavelets while simultaneously capturing the positivity and "spikiness" of nonGuassian traffic. We develop a queuing analysis for our model by exploiting its multiscale construction scheme. We elucidate our model's ability to capture the covariance structure of real data and then fit it to real traffic traces. Queuing experiments demonstrate the accuracy of the model for matching real data and the precision of our theretical queuing result. Our results indicate that a Guassian assumption can lead to over-optimistic predictions of tail queue probability.