Database Monte Carlo for simulation on demand

In the simulation-on-demand paradigm, we invest computational effort by running a simulation experiment before a question is asked, and then we quickly provide an answer by making use of the results of the earlier simulation experiment. This can be done by building a metamodel, but standard metamodeling methods used in stochastic simulation have the disadvantage that they require validation. We show how to use Database Monte Carlo with control variates to provide simulation-on-demand without metamodel validation.