An asymptotically optimal greedy algorithm for large optical burst switching systems

Lachlan L.H. Andrew, Yuliy Baryshnikov, E. G. Coffman, Stephen V. Hanly, Jolyon White

Research output: Contribution to journalArticlepeer-review

Abstract

As the number of wavelengths in OBS systems increases, the utilization achievable for a given blocking probability can be made to approach 100%. This paper shows that this property applies to a wavelength allocation algorithm of greedy type. Another property of this rule, one shared by most other wavelength assignment algorithms, is that, since lost traffic tends to occur near destinations, where the resource usage wasted by such traffic is large, very low blocking probabilities are important for efficient operation. To help identify regions of low blocking probability, we derive an asymptotically exact condition for zero blocking probabilities; it has a form reminiscent of the stability condition of the M/G/1 queue.

Original languageEnglish (US)
Pages (from-to)14-16
Number of pages3
JournalPerformance Evaluation Review
Volume31
Issue number2
DOIs
StatePublished - Sep 2003
Externally publishedYes

Keywords

  • Fluid limits
  • Hydrodynamic limits
  • Optical burst switching
  • Optical networks
  • Stochastic modeling
  • Wavelength division multiplexing

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Computer Networks and Communications

Fingerprint

Dive into the research topics of 'An asymptotically optimal greedy algorithm for large optical burst switching systems'. Together they form a unique fingerprint.

Cite this