This paper examines a problem of importance to the telecommunications indus
try. In the design of modern ATM switches, it is necessary to use simulatio
n to estimate the probability that a queue within the switch exceeds a give
n large value. Since these are extremely small probabilities, importance sa
mpling methods mast be used. Here we obtain a change of measure for a broad
class of models with direct applicability to ATM switches.
We consider a model with A independent sources of cells where each source i
s modeled by a Markov renewal point process with batch arrivals. We do not
assume the sources are necessarily identically distributed, nor that batch
sizes are independent of thr state of the Markov process. These arrivals jo
in a queue served by multiple independent servers, each with service times
also modeled as a Markov renewal process. We only discuss a time-slotted sy
stem. The queue is viewed as the additive component of a Markov additive ch
ain subject to the constraint that the additive component remains non-negat
ive. We apply the theory in McDonald (1999) to obtain the asymptotics of th
e tail of the distribution of the queue size in steady state plus the asymp
totics of the mean time between large deviations of the queue size.