An Algorithmic Toolbox for Network Calculus

  • Authors:
  • Anne Bouillard;Éric Thierry

  • Affiliations:
  • IRISA/ENS Cachan (Bretagne), Bruz, France 35170;LIP (ENS Lyon, CNRS UMR 5668, INRIA, UCBL), Lyon Cedex 07, France 69364

  • Venue:
  • Discrete Event Dynamic Systems
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Network calculus offers powerful tools to analyze theperformances in communication networks, in particular to obtaindeterministic bounds. This theory is based on a strong mathematicalground, notably by the use of (min,+) algebra. However, thealgorithmic aspects of this theory have not been much addressedyet. This paper is an attempt to provide some efficient algorithmsimplementing network calculus operations for some classicalfunctions. Some functions which are often used are the piecewiseaffine functions which ultimately have a constant growth. As afirst step towards algorithmic design, we present a classcontaining these functions and closed under the main networkcalculus operations (min, max, +, -, convolution, subadditiveclosure, deconvolution): the piecewise affine functions which areultimately pseudo-periodic. They can be finitely described, whichenables us to propose some algorithms for each of the networkcalculus operations. We finally analyze their computationalcomplexity.