Management of real-time quality of service using (m,k)-firm constraints into packet-switched networks
Gestion de la qualité de service temporelle selon la contrainte (m,k)-firm dans les réseaux à commutation de paquets
Abstract
This work focuses on the definition of a novel approach of real-time QoS using the (m,k)-firm constraints into packet switched networks. We define the concept of loss-tolerant QoS with graceful degradation that consists in guaranteeing an acceptable level of QoS for loss-tolerant real-time applications in overload situation. Our approach is mainly applied to multimedia networks assuring the delivery of real-time loss-tolerant multimedia flows such as MPEG video broadcast, Voice over Ir, audio streams ... The contribution presented in this work is (1) to propose a novel scheduling algorithm to provide loss-tolerant QoS with graceful degradation, (2) validate the proposai through different and complementary techniques: mathematical analysis, simulation and implementation in test bed platform. Our approach of gracefully degraded QoS is based two concepts: (1) Bandwidth guarantee, (2) meeting the (m,k)-firm timing constraints. We have combined both concept to propose our scheduling algorithm, called (m,k)-WFQ, that performs fair resources sharing as dos WFQ while taking into account the (m,k)-firm timing constraints of each stream.
Le cadre d'application de ce travail vise principalement les réseaux assurant le transport des flux temps-réel tolérant les pertes telles que la voix sur Ir, la transmission de la vidéo MPEG, les flux audio ... L'apport principal de cette thèse réside dans la proposition dun nouvel algorithme d'ordonnancement pour fournir une QdS acceptablement dégradable et la validation de cette proposition en combinant les techniques de vérification: analyse mathématique, simulation et implémentation sur une plateforme. Notre approche de QdS acceptablement dégradable est basée sur deux concepts: (1) la garantie de bande passante (WFQ), (2) le respect de la contrainte temporelle (m,k)-firm. Nous avons combiné ces deux notions pour proposer l'algorithme d'ordonnancement (m,k)-WFQ assurant un partage équitable des ressources et respectant la contrainte (m,k)-firm de chaque flux.