A relaxation time approach for reducing initialization bias in simulation
ANSS '85 Proceedings of the 18th annual symposium on Simulation
Simulation Modeling and Analysis
Simulation Modeling and Analysis
Steady-state simulation of queueing processes: survey of problems and solutions
ACM Computing Surveys (CSUR)
| Hi-index | 0.00 |
In a previous paper, a heuristic technique was introduced for reducing initialization bias in Monte-Carlo simulation of stationary, infinite-capacity, Markovian queueing systems. The technique differs from many existing methods in that, rather than relying on a statistical analysis of output from preliminary simulation runs, it is based on characteristics of the underlying theoretical model. Experimental results presented earlier provide convincing evidence that use of the heuristic yields an unbiased estimate of the steady-state expected queue length in M/M/1 systems with virtually no cost added to the simulation. We now present results of experiments designed to validate the technique for estimating the steady-state expected number of customers in M/Ek/1 queueing systems.