Title :
Long-range dependence does not necessarily imply non-exponential tails
Author :
Makowski, Armand M.
Author_Institution :
Dept. of Electr. & Comput. Eng., Maryland Univ., College Park, MD, USA
Abstract :
We exhibit a buffer model with a provably long-range dependent input source but whose corresponding measure of buffer occupancy is exponentially distributed (and thus has exponential tails). This example invalidates the perception which has emerged in the literature on traffic engineering, that long-range dependencies necessarily induce nonexponential tails for the buffer asymptotics.
Keywords :
buffer storage; statistical analysis; telecommunication traffic; buffer asymptotics; buffer model; buffer occupancy; exponential tails; long-range dependence; long-range dependent input source; on-off fluid models; traffic engineering; Asynchronous transfer mode; Brownian motion; Ethernet networks; Predictive models; Probability distribution; Tail; Telecommunication traffic; Time measurement; Traffic control; Wide area networks;
Journal_Title :
Communications Letters, IEEE
DOI :
10.1109/LCOMM.2002.806452