Tail Transitions in Queues with Long Range Dependent Input

  • Authors:
  • Tim Daniëls;Chris Blondia

  • Affiliations:
  • -;-

  • Venue:
  • NETWORKING '00 Proceedings of the IFIP-TC6 / European Commission International Conference on Broadband Communications, High Performance Networking, and Performance of Communication Networks
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper studies the tail of the buffer occupancy distribution of a queueing system with c parallel deterministic servers, infinite buffer capacity and an input process with consists of a superposition of a long range dependent on-off source and a batch renewal process. In particular, we investigate the decay of the tail for different values of c and for different compositions of the traffic mix. It is shown that for c = 1 (i.e. a single server system), the tail has a power law decay, while for c 1, different cases may be distinguished: if the arrival rate of the background traffic is larger than c - 1, then the tail probabilities decay according to a power law, while for the other case, the decay is exponential.