On the application of accelerating simulation methods in network analysis

When evaluating quantitative aspects of communication networks using simulation, one of the main difficulties to face is the often considerable computing power required. In some specific mathematical frameworks, different techniques have been proposed in order to accelerate the simulation process while keeping the same precision in the result. This paper deals with one of them, the so called importance splitting technique, applied to a general discrete event simulator used to estimate low probability values, namely loss probabilities. We present experimental evidence that this approach can be efficient and we point out some of the difficulties that the user can find in following it. We also show that this method can result in accelerating factors even if the estimated probabilities are not very low. KEY TOPICS: Modeling and simulation of communication systems. Submitted for the General Conference J. Incera, G. Rubino, N. Stier, “On the application of accelerating simulation methods in network analysis” 1 On the application of accelerating simulation methods in network analysis José Incera y ENST Bretagne 2 Rue de la Châtaigneraie 35512 Cesson Sévigné, France [email protected] Gerardo Rubino ENST Bretagne and IRISA Campus Beaulieu 35042 Rennes Cedex, France [email protected] Nicolas Stier IRISA Campus Beaulieu 35042 Rennes Cedex, France