HEAF: A Novel Estimator for Long-Range Dependent Self-similar Network Traffic

Long-range dependent (LRD) self-similar chaotic behaviour has been found to be present in internet traffic by many researchers. The ‘Hurst exponent’, H, is used as a measure of the degree of long-range dependence. A variety of techniques exist for estimating the Hurst exponent; these deliver a variable efficacy of estimation. Whilst ways of exploiting these techniques for control and optimization of traffic in real systems are still to be discovered, there is need for a reliable estimator which will characterise the traffic. This paper uses simulation to compare established estimators and introduces a new estimator, HEAF, a ‘Hurst Exponent Autocorrelation Function’ estimator. It is demonstrated that HEAF(2), based on the sample autocorrelation of lag2, yields an estimator which behaves well in terms of bias and mean square error, for both fractional Gaussian and FARIMA processes. Properties of various estimators are investigated and HEAF(2) is shown to have promising results. The performance of the estimators is illustrated by experiments with MPEG/Video traces.