IS simulation for the waiting time of ATM burst cell traffic

Abstract

AbstractInvestigations of cell delay characteristics for various kinds of traffic are necessary for the design of buffer sizes in ATM converters. The purpose of this research is to obtain the cell delay characteristics, particularly the cell delay α value, using simulations. However, using the prior method, the Monte Carlo (MC) method, requires considerable design time and imposes limits on the precision of the pseudo‐random numbers. As a result, cell delay characteristics for a very small α cannot be obtained. Therefore, the purpose of this research is to apply Importance Sampling (IS) and obtain cell delay characteristics quickly at a precision below 10−9. In particular, the IS method is applied to the waiting time characteristics in the steady state for Geo/D/1, a queue model for random cell traffic, and to the two‐state MMGP/D/1 queue for burst cell traffic. An approximate optimal simulation distribution is obtained based on large deviation theory. Then an actual IS simulation is performed using the obtained simulation distribution. Compared to the MC simulation, the IS simulation is shown to have superior speed and a higher confidence interval. In addition, the IS estimated values are shown to be more precise compared to theoretical values or approximate values. © 2002 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 85(11): 97–105, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.1142