Abstract
This paper considers a two-station system in which the second station has a finite buffer size and the discipline is to drop or block customers when it is full. The dropping discipline is likely to be encountered in broadband ISDN systems carrying video traffic. Semi-Markovian arrivals are allowed as the input to the tandem queues which can be used to capture the effect of correlations in the input process. The solution consists of studying the stochastic behavior of the attained waiting time of the customer in service in the first station. From this result, one can obtain sojourn-time and queue-length distributions and the probability that a customer is dropped. The method is illustrated by numerical examples.