Pr. Jelenkovic et al., THE EFFECT OF MULTIPLE TIME SCALES AND SUBEXPONENTIALITY IN MPEG VIDEO STREAMS ON QUEUING BEHAVIOR, IEEE journal on selected areas in communications, 15(6), 1997, pp. 1052-1071
Guided by the empirical observation that realtime MPEG video streams e
xhibit both multiple time scale and subexponential characteristics, we
construct a video model that captures both of these characteristics a
nd is amenable to queueing analysis, We investigate two fundamental ap
proaches for extracting the model parameters: using sample path and se
cond-order statistics-based methods. The model exhibits the following
two canonical queueing behaviors. When strict stability conditions are
satisfied, i.e., the conditional mean of each scene is smaller than t
he capacity of the server, precise modeling of the interscene dynamics
(long-term dependency) is not essential for the accurate prediction o
f small to moderately large queue sizes, In this case, the queue lengt
h distribution is determined using quasistationary (perturbation theor
y) analysis. When weak stability conditions are satisfied, i.e., the c
onditional mean of at least one scene type is greater than the capacit
y of the server, the dominant effect for building a large queue size i
s the subexponential (long-tailed) scene length distribution, In this
case, precise modeling of intrascene statistics is of secondary import
ance for predicting the large queueing behavior, A fluid model, whose
arrival process is obtained from the video data by replacing scene sta
tistics with their means, is shown to asymptotically, converge to the
exact queue distribution. Using the transition scenario of moving from
one stability region to the other by a change in the value of the ser
ver capacity, we synthesize recent queueing theoretic advances and ad
hoc results in video modeling, and unify a broad range of seemingly co
ntradictory experimental observations found in the literature, As a wo
rd of caution for the widespread usage of second-order statistics mode
ling methods, we construct two processes with the same second-order st
atistics that produce distinctly different queueing behaviors.