Performance evaluation of dependent two-stage services

In many scenarios services are provided in successive stages. While tandem queues appropriately reflect the structure of such scenarios, the typical assumption that service times at different stages are independent often does not fit to reality. We examine, via simulation, the impact of dependencies among service times on expected customer waiting times. The usual network simulation overhead caused by event list handling is avoided by an extension of the Lindley recursion to the two-stage case. Numerical results are presented for exponentially and uniformly distributed service times with different types of dependencies and varying server utilizations.