A study on efficiency improvement of a teletraffic performance simulator using variance reduction techniques

This study focuses on the application of variance reduction techniques to improve the efficiency of a conventional teletraffic performance evaluation simulator in the context of Monte Carlo simulation of a T-T digital switching system. Antithetic variable and control variable techniques were used and several implementations were considered, for calculating the confidence intervals. For reference and comparison of the various implementations the authors considered the calculation of the confidence interval by the method of the batch means. A major conclusion is that for tight relative confidence intervals (5%-10%) the use of the mean of the (Poissonian) traffic offered, as control variable, is most advantageous.