Fractional noise in experimental measurements of IP traffic in a metropolitan area network

Internet traffic exhibits self-similarity and long-range dependence (LRD) on various time scales. In this work, we analyze IP traffic traces measured on a MAN link of the telecommunications operator FASTWEB. We emphasize results obtained using the modified Allan variance (MAVAR), a well-known time-domain tool originally conceived to discriminate noise types with power-law spectrum (i.e., fractional noise) in frequency stability measurement, which has been introduced in network traffic analysis only recently. MAVAR has been demonstrated to feature superior spectral sensitivity and accuracy in fractional-noise parameter estimation, coupled with outstanding robustness against nonstationarity in data analyzed. Logscale diagrams are also evaluated, to confirm MAVAR experimental findings. All traffic traces analyzed exhibit prominent evidence of fractional noise. Model parameters are estimated. Results obtained are valuable to establish IP traffic models, e.g. necessary for realistic input traffic synthesis in network simulation