Philippe Olivier

Dimensioning high speed IP access networks

This paper discusses the definition of simple dimensioning rules for high speed IP access networks carrying data traffic. We notably provide formulas relating capacity, demand and performance allowing dimensioning for a target quality of service expressed in terms of useful per-flow throughput. These formulas derive from a data traffic model equivalent of the Engset model for telephone access networks. Performance is shown to be largely independent of precise traffic characteristics. The key dimensioning parameter is offered traffic defined as the average data rate a user would generate in the absence of congestion.

Flow level IP traffic characterization

In this paper we perform a detailed statistical analysis of IP traffic at flow level, based on data traces collected on the France Telecom wide area IP network. We particularly focus on the assessment of assumptions commonly used in flow level performance models, i. e. Poisson flow arrival process, both for UDP and TCP traffic. We first discuss the concept of flow and present a method to estimate an adequate Time Out value to distinguish different flows within a traffic stream. To model the flow arrival process, we find that Gamma and Weibull distributions provide excellent fits to the empirical flow inter-arrival time distributions. In addition, a fairly low level of correlation is detected between successive flow arrivals. Flow size distributions, though more difficult to model due to their composite aspect, show a very well marked heavy-tailed character of the Pareto type. We found that these results were very consistent among all the data traces we analyzed.

Internet data flow characterization and bandwidth sharing modelling

The purpose of the paper is two-fold: first, characterize the Internet data traffic at high levels, i.e., flows and sessions, with the principal aim of comparing the salient features shown by different types of applications, namely P2P on one hand and classical Internet applications in the other. Second, among the observed traffic characteristics, we particularly focus on the way bandwidth is actually shared by flows and users by developing an empirical bandwidth sharing model. We infer from it some trends shown by the different classes of traffic in terms of elasticity, i.e., their potential reaction against congestion.

Dimensioning IP access links carrying data traffic

This paper proposes simple dimensioning rules for high speedip access links carrying data traffic. Assuming a finite source population and fair bandwidth sharing among user flows, we derive formulas relating capacity, demand and performance. These formulas allow link dimensioning for a target quality of service expressed in terms of useful per-flow throughput. They constitute a data traffic model equivalent of the Engset model for telephone access networks. Performance is shown to be largely independent of detailed traffic characteristics such as the statistical distributions of flow size and think time. Simple approximations are derived for two distinct performance regimes corresponding to transparency and saturation, respectively. Extensions to the basic model account for a heterogeneous user demand, unfair bandwidth sharing or different access rate classes. In any case, the key parameter for dimensioning is the offered traffic, defined as the average data rate a user would generate in the absence of congestion.RésuméCet article propose des règles de dimensionnement simples applicables à des liaisons d’accèsip haut débit transportant du trafic de données. Considérant une population source de taille finie et sous l’hypothèse d’un partage équitable de la bande passante par les flots des utilisateurs, on obtient les formules reliant la capacité, la demande et la performance. Ces formules permettent le dimensionnement des liaisons en fonction d’un objectif de qualité de service s’exprimant en termes de débit utile de transfert par flot. Elles constituent un modèle de trafic de données équivalent au modèle d’Engset utilisé pour les réseaux d’accès téléphoniques. On montre que la performance est dans une large mesure indépendante des caractéristiques détaillées du trafic, telles que les lois statistiques de la taille des flots ou de la durée des périodes de silence. Des approximations simples de la performance sont obtenues pour deux régimes de fonctionnement opposés, les régimes transparent et saturé. Des extensions du modèle de base sont proposées, prenant en compte les possibilités d’une demande hétérogène des utilisateurs, d’un partage inéquitable de la bande passante, ou d’une segmentation en classes de débits d’accès différents. Le paramètre clé pour le dimensionnement est dans tous les cas le trafic offert, défini comme le débit moyen de transfert qu’un utilisateur engendrerait en l’absence de congestion.

Performance Evaluation of Multi-rate Streaming Traffic by Quasi-Stationary Modelling

A Quasi-Stationary approach to evaluate streaming traffic performance was recently proposed, which combines in a simple way the traffic variations at flow and packet time-scales. The present paper elaborates on it, by specifically focusing on the multi-class/multi-rate aspects of the model. A first original contribution is the theoretical description of flow peak rates in terms of a continuous distribution. Then, a set of simulations is reported which shows in some simple cases, e.g., a multiplex of constant bit rate flow classes, the suitability and efficiency of the approach. Finally, a preliminary trace-driven simulation experiment, although not fully satisfactory, shows some need for explicitly taking into account the flow peak rates in performance evaluation.

IP-Based Networks

Contributors: Mine Caglar (Koc University, Turkey), Ivan Gojmerac (FTW Vienna, Austria), Martin Klimo (University of Zilina, Slovakia), Michael Menth (University of Wurzburg, Germany), Michela Meo (Politecnico di Torino, Italy), Sandor Molnar (Budapest University of Technology and Economics, Hungary), Ilkka Norros (VTT, Finland), Philippe Olivier (France Telecom, France), Sara Oueslati (France Telecom, France), Oznur Ozkasap (Koc University, Turkey), Halina Tarasiuk (Warsaw University of Technology, Poland)

A macroscopic overall grade of service model for telephone circuit- switched networks

Cet article presente un modele theorique analysant la performance globale d’un reseau telephonique, exprimee comme la distribution des taux de blocage par flux, en fonction de regles de dimensionnement appliquees localement aux faisceaux de circuits du reseau. Le modele traite explicitement les parties de reseau fonctionnant en acheminement hierarchique, et peut integrer des distributions de blocage obtenues de facon exogene, et portant sur des parties de reseau non modelisees. Il donne une estimation probabiliste de la qualite de service de bout en bout percue par l’ensemble des usagers, un jour quelconque de l’annee. Ce modele peut etre ap-plique a de grands reseaux, et constitue dans ce cas, un outil theorique pour determiner les caracteristiques d’une gestion du trafic par destination.This paper presents an analytical model allowing the analysis of the overall grade of service of a network expressed in terms of stream blocking distribution, as a function of dimensioning rules applied at a local level. The model applies directly to parts of the network working under fixed hierarchical routing, and can integrate blocking distributions obtained by exogenous means, and concerning parts of the network not explicitly modeled. The model estimates in a probabilistic way the end- to- end grade of service as perceived by the customer any day of the year. This model can be applied to large networks, and, in such case, is an efficient theoretical tool to determine the characteristics of a stream oriented traffic management.RésuméCet article présente un modèle théorique analysant la performance globale d’un réseau téléphonique, exprimée comme la distribution des taux de blocage par flux, en fonction de règles de dimensionnement appliquées localement aux faisceaux de circuits du réseau. Le modèle traite explicitement les parties de réseau fonctionnant en acheminement hiérarchique, et peut intégrer des distributions de blocage obtenues de façon exogène, et portant sur des parties de réseau non modélisées. Il donne une estimation probabiliste de la qualité de service de bout en bout perçue par l’ensemble des usagers, un jour quelconque de l’année. Ce modèle peut être ap-pliqué à de grands réseaux, et constitue dans ce cas, un outil théorique pour déterminer les caractéristiques d’une gestion du trafic par destination.