Date Added: Feb 2011
Characterizing backbone networks poses a significant challenge due to the unstable and fluctuated behavior exhibited by network traffic dynamics. Modeling techniques developed for volume-based traffic profiling rely on the statistical assumptions of stationarity, Gaussianity and linearity, whose applicability has not been thoroughly investigated throughout past and recent work. The authors argue that modeling assumptions should be rigorously validated since they determine the accuracy of any model applied to describe the traffic process. In this paper, they introduce and illustrate the suitability of Time-Frequency (TF) representations and the Hinich algorithms for the validation of modeling assumptions on captured backbone and edge link network traces.