Operating system efficiency evaluation on the base of measurements analysis with the use of non-extensive statistics elements

Abstract

The major goal of this article was to evaluate the efficiency of Linux operating systemusing statistical self-similarity and multifractal analysis. In order to collect the necessary data, thetools available in Linux such as vmstat, top and iostat were used. The measurement data collected withthose tools had to be converted into a format acceptable by applications which analyze statistical selfsimilarityand multifractal spectra. Measurements collected while using the MySQL database systemin a host operating system were therefore analyzed with the use of statistical self-similarity and allowedto determine the occurrence of long-range dependencies. Those dependencies were analyzed with theuse of adequately graduated diagrams. Multifractal analysis was conducted with the help of FracLabapplication. Two methods were applied to determine the multifractal spectra. The obtained spectrawere analyzed in order to establish the multifractal dependencies.