When, and when not to Use Warm-up Periods in Discrete Event Simulation

Winfried K. Grassmann
Professor Emeritus
Dept. of Computer Science, University of Saskatchewan

Abstract

This paper demonstrates, both theoretically and by using numerical examples, that if one has good starting states, one should not use warm up periods in discrete event simulation when estimating equilibrium expectations by means of time averages. The numerical methods used are deterministic, and are based on randomization or uniformization. We also point out the main differences between deterministic and stochastic equilibria, and we show that there are hybrid systems, which require warm up periods.

Biographical Notes

Winfried Grassmann joined the University of Saskatchewan in 1968, after working in the airline industry as an Operations Research Analyst for 4 years. He has published several books and numerous papers. He is best known for the randomization method and the GTH algorithm, methods to find transient and steady-state solutions for Markovian event systems, such as queueing networks. Both algorithms have been implemented by several institutions in in the USA and Europe in commercial software packages.

April 14, 2009

Park Town Hotel

Cocktails 5:30, Dinner 6:00, Presentation 7:00

Tickets $20

For more information contact:

Kent Kostuk 244-3295 k.kostuk@fcl.ca

Winfried Grassmann, 966-4898 grassman@cs.usask.ca