ABOUT METHODS FOR ANALYSIS OF SELFSIMILARITY OF NETWORK TRAFFIC

Abstract

The article research networks traffic self-similarity analysis methods. Applications of Hurst
statistics for calculation of Hurst coefficient, frequency/wave features estimators methods -
Periodograms, Whittle, Abby-Weich have been analyzed. Suitability of employed methods for
analysis was tested by the way of computer-based simulation. The analyzed methods has been
tested by applying Fractan programme’s R/S statistics, Selfis programme by applying time
analysis and frequency/wave features’ estimation methods. R. Weron’s (2004) algorithm of
generation of random standard stable values has been used for forming self-similar network
traffic time series, where stability index α=1.8 (H=0.56). The results obtained with Fractan and
Selfis show that Hurst coefficient changes from 0.53 to 0.70, the stability index changes from
0.53 to 1.89.
Key words: self-similarity, Hurst coefficient, α-stable distribution, traffic burstiness.