A Tool for Long-Range Dependent Analysis via the R/S Statistic

Self-similarity and long-range dependence have been found to apply as models of traffic in modern computer networks. This behavior has important implications for the design, analysis, control and performance of such networks, thus an accurate identification and quantification of such behavior is important. A novel tool for the analysis of longrange dependence via the R/S statistic is presented. The tool, R/S Analyzer, facilitates the analysis of LRD traces by providing a two-step approach for estimating the Hurst parameter. The accuracy of the tool is tested by using several synthetic and real long-range dependent traces with known Hurst parameter. A comparison of the tool against similar tools for long-range dependence analysis is performed. We show that our tool presents better estimates of the Hurst parameter and behaves better against aggregation in time.