Simulating traffic for an ATM network may be very time consuming due to the large amount of transmitted cells, if traditional simulation modeling at cell level is employed. A possible means to speed up the simulations is to model the traffic at a higher abstraction level. In this paper, we evaluate different simulation modeling techniques on a scheme for call admission control in ATM networks. By generating simulation events only at the start and end of each constant-rate traffic burst, up to 85% reduction of the simulation time was obtained, while keeping the validity of the simulation model. This opens new possibilities to simulate large and complex networks with large volumes of high-speed traffic. In particular, this leads us one step closer to simulating network performance in real time as a support to network management functions.