Insensibilité et bornes stochastiques dans les réseaux de files d'attente. Application à la modélisation des réseaux de télécommunication au niveau flot.

Abstract : The thesis deals with the performance evaluation of multi-service telecommunication networks. We first concentrate the analysis on data traffic performance: using the theory of Processor Sharing queueing networks, we study the different possible bandwidth allocations. In particular, we identify the allocations leading to a mean flow-level performance that is insensitive to detailed statistical traffic characteristics. Then, we try to quantify end-to-end packet delay for real-time applications and give, using the notion of stochastic ordering, arguments in favour of the so-called "better than Poisson" conjecture leading to a tight upper bound of the delay. Finally, we study the impact of real-time traffic on data traffic performance. This analysis is also based on some advanced stochastic ordering tools.
Document type :
Theses
Complete list of metadatas

https://pastel.archives-ouvertes.fr/pastel-00000733
Contributor : Ecole Polytechnique <>
Submitted on : Wednesday, July 21, 2010 - 10:30:46 AM
Last modification on : Wednesday, July 28, 2010 - 10:13:36 AM
Long-term archiving on : Friday, October 22, 2010 - 3:36:42 PM

Identifiers

  • HAL Id : pastel-00000733, version 1

Collections

Citation

Alexandre Proutière. Insensibilité et bornes stochastiques dans les réseaux de files d'attente. Application à la modélisation des réseaux de télécommunication au niveau flot.. Mathématiques [math]. Ecole Polytechnique X, 2003. Français. ⟨pastel-00000733⟩

Share

Metrics

Record views

383

Files downloads

314