On the efficient simulation of large communication networks using importance sampling

To obtain large speedup factors in Monte Carlo simulation using importance sampling (IS), the modification, or bias of the underlying probability measures must be carefully chosen, especially for large networks. The formulation and results for estimating the blocking probability of a 4×4 synchronous Clos ATM switch, with priority and push-out, are presented. The simulation study uses the authors´ (1992) statistically-based techniques for optimizing IS performance. Issues involved in using IS that, if not properly addressed, can result in poor performance and even misleading results are presented. Techniques are presented to avoid these potential pitfalls. Experimental examples of both the 4×4 Clos ATM switch and IS-related issues are presented. Speedup factors of 2 to 5 orders of magnitude are achieved for the 4×4 ATM switch