Characterization and multiplexing of correlated traffic from sources with different time constants

  • Authors:
  • Ioannis Stavrakakis

  • Affiliations:
  • Electrical and Computer Engineering, 409 Dana Building, Northeastern University, Boston, MA 02115, USA

  • Venue:
  • Computer Communications
  • Year:
  • 1996

Quantified Score

Hi-index 0.24

Visualization

Abstract

Single server, discrete-time queueing systems under independent or Markov arrival processes and deterministic service requirements have been studied widely in the past. The fixed service time is usually set equal to one (the network-time-constant), defining the time unit of the discrete-time system axis. Such queueing systems have been adopted for the modeling of statistical multiplexers in slotted communication networks supporting sources of (fixed size) packetized information. In this paper, a statistical multiplexer fed by Markov packet sources is considered. Unlike previous considerations, the time-constant of the sources is assumed to be greater than one. As a result, packets cannot be generated over consecutive discrete-time instants whose distance is less than the source time-constant, although the source may be active; both the cases of Markov sources with single and double time-constants are considered. The resulting packet traffic is characterized and the formulated queueing model is studied. In addition to presenting an analysis of a statistical multiplexer fed by quite general Markov traffic streams, this study concludes that the negative impact of source traffic correlations on statistical multiplexing can be reduced substantially when sources with time-constants greater than one are considered. Even in the absence of correlations, the guaranteed idleness of the source within a source-time-constant's interval is shown to have a positive impact on the multiplexing process. Finally, some numerical results for the case of multiplexed packetized voice sources are also presented.