We propose a call admission control (CAC) procedure for asynchronous t
ransfer mode (ATM) networks using statistical bandwidth as the decisio
n criterion, based on closed-form expressions that use diffusion model
s. This approach is computationally very efficient and easily implemen
table. It is related to Gaussian approximations previously proposed fo
r CAC, though our expressions are derived from a more derailed represe
ntation of traffic and the buffer length process, The statistical band
width expressions we use take into consideration the users' cell loss
requirements, their aggregate traffic characteristics, the available b
uffer size at the statistical multiplexers, and capture the interactio
n between individual traffic streams at the ATM multiplexer. Extensive
numerical and simulation results are presented to evaluate the effici
ency and adequacy of this CAC procedure. Comparison with existing meth
ods such as the Equivalent Bandwidths and Gaussian Approximations indi
cate that our approach remains conservative with respect to cell loss
compared to previously proposed CAC schemes, yet that it is more econo
mical in bandwidth allocation leading to larger admission regions both
for homogeneous and heterogeneous traffic.