Long range correlation in multiplexed pareto traffic

Packet traffic from various sources - ethernet, ISDN, CCSN and VBR video - has been shown to exhibit self-similarity and related properties of long range correlation and slowly decaying variances. The author has previously suggested that self-similar traffic can be modelled by heavy-tailed pareto sources. A traffic stream in which interarrival times are iid pareto generates block packet counts which are asymptotically self-similar. However, single pareto streams do not produce the power law or stretched exponential queue tails observed with real packet traces. This paper shows that multiplexed streams of pareto traffic exhibit long range correlation of block packet counts and interarrival times, and that these properties together produce power law queue tails. This result makes pareto sources an attractive framework for modeling self-similar traffic.