Recurrence and regeneration in non-markovian networks of queues

Closed networks of queues with priorities among job classes provide a framework for modeling a wide variety of complex congestion phenomena. Steady state estimation procedures based on underlying regenerative process structure are available only for non-Markovian networks in which all service interruptions are of the preempt-repeat type and all service time distributions have positive density on the positive half-line. Using “new better than used” distributional assumptions and sample path structure of the underlying job stack process, we establish criteria for recurrence and regeneration that avoid these restrictions. Regenerative simulation methods follow from these results.