Asynchronous transfer mode (ATM) is the transmission and switching sys
tem that has been adopted for next generation multimedia networks. It
is based on statistical multiplexing techniques and the performance of
an ATM based network is quantified by the quality of service, QoS, de
livered to the network users. QoS is defined in terms of the maximum c
ell loss rate and the delay/jitter allowable for a given network servi
ce connection. Normally, network performance is evaluated through the
use of either mathematical or computer simulation modelling. The two m
odel types differ in their individual limitations and the extent to wh
ich they can represent the real system. In this application, a compute
r simulation model has the advantage of flexibility and accuracy but i
s limited by the extent to which cell loss rate can be calculated. For
video services, for example, a cell loss rate of 1E-11 is specified w
hich would require an unacceptably high amount fo computer simulation
time to investigate. In order to investigate such issues, therefore, t
his paper describes an approach which integrates both mathematical and
computer simulation modelling. Simulation results are presented to pr
ove the validity of the proposed approach, which is applicable for use
in other application environments involving rare-event occurrences.