Tail probabilities for the multiplexing of fractional α-stable broadband traffic

We investigate the tail probabilities of a multiplexer driven by α-stable self-similar traffic. We consider a parsimonious 4-parameter traffic model, which best captures the long range dependence and heavy-tails of aggregate packet traffic in broadband networks. Input traffic with these characteristics, induces buffer dynamics which are qualitatively different from those which arise in traditional traffic management. Using the effective bandwidth theory, we extend the recent results on α-stable self similar-driven queues with infinite buffer to the finite buffer case that model routers/switches more accurately. Queuing simulations with real broadband traffic emphasise the improvements in engineering considerations (e.g., connection admission control, buffer management, statistical multiplexing gains), with respect to the existing results so far. Our experiments involve a large set of real broadband network traffic which consists of Ethernet LAN, Internet WAN and MPEG-1 compressed video traces