Using Augmented Infinitesimal Perturbation Analysis forCapacity Planning in Intree ATM Networks

  • Authors:
  • Charles A. Brooks;Pravin Varaiya

  • Affiliations:
  • Department of Electrical Engineering and Computer Science, University of Illinois at Chicago, 851 S. Morgan Street, Chicago, IL 60607-7053;Department of Electrical Engineering and Computer Sciences, University of California at Berkeley

  • Venue:
  • Discrete Event Dynamic Systems
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Augmented Infinitesimal Perturbation Analysis is usedto determine asymptotically unbiased and strong consistent gradientestimates for use in the capacity planning of intree ATM networks.These gradients are used to determine the locally optimal minimumaverage network delay by applying a steepest descent algorithmwith projection and an Armijo line search to solve the capacityassignment (CA) problem. The network capacities are governedby a linear cost constraint. It is assumed that input regulators(e.g., leaky bucket regulators) are used at the source of eachvirtual circuit. All virtual circuit external arrivals are modeledas independent Poisson processes. Regenerative simulation isused to determine gradient estimates.