Ordinal optimization by means of standard clock simulation and crude analytical models

In this paper we apply the standard clock (SC) and ordinal-optimization techniques to the voice-call admission-control problem in integrated networks, which is a problem of considerable interest to the communications community. We actually had to extend the SC model to incorporate deterministic events, namely the fixed packet lengths of data packets. Our ordinal-optimization studies show that crude models are often adequate to predict the relative performance of different control policies, even though they may provide poor estimates of actual system performance. The accuracy of these rankings suggests that simple analytical models can be used to reduce the search space to just a few policies whose performance can then be evaluated accurately via simulation, thus decreasing computation time dramatically