This paper presents a source model for VER video traffic, A finite-state Ma
rkov chain is shown to accurately model one- and two-layer video of all act
ivity levels on a per source basis. Our model captures the source dynamics,
including the shortterm correlations essential for studying network perfor
mance. The modeling technique is shown to be applicable for both H.261 and
MPEG-2 encoded video of a variety of activity levels. The traffic model is
shown in a simulation study to be able to accurately characterize both the
single-source buffer occupancy over a nide range of buffer sizes and the mu
ltiplexing behavior. The VER video model is also used to model the enhancem
ent layer of two-layer SNR scalable video. We show that two-layer encoding
has significantly better statistical multiplexing gains than one-layer vide
o, particularly when the network admits calls based on a leaky-bucket chara
cterization.