Performance Analysis of Time-domain Algorithms for Self-similar Traffi

Self-similarity has been found in Internet traffic, however there is still certain confusion in the way techniques for detecting Hurst exponent work and how are these affected by certain parameters. In this paper, we identify these parameters showing the impact they have in the final estimation of the Hurst parameter. We provide a deep study of these tuning parameters and show the optimum values. Aditionally we present the Hurst Estimator of a tool called SelQoS and compare its time-domain techniques with those of the tool called Selfis. We show that SelQoS gives more accurate estimations of the Hurst exponent and that Selfis tends to underestimate the Hurst exponent when using real and synthetic self-similar traffic traces.