soutenance de thèse de Yuhui Wang,

« De l’utilisation du codage reseau et du multicast pour améliorer la performance dans les réseaux filaires »,

le Vendredi 17 Mai 2013 à 11h00 à l’Université Pierre et Marie Curie – Paris 6, salle 25-26/105, à Jussieu.

Le jury sera composé de :

– Rapporteur : Fabio Martignon, Professeur, Universite Paris-Sud

– Rapporteur : Dritan Nace, Professeur, Universite de Technologies de Compiegne

– Examinateur : Prosper Chemouil, Directeur de Recherche, Orange Labs

– Examinateur : Philippe Chretienne, Professeur, Universite Pierre et Marie Curie

– Examinateur : Muriel Medard, Professeur, Massachusetts Institute of Technology

– Encadrant de these : Eric Gourdin, Ingenieur de Recherche, Orange Labs

– Directeur de these : Eitan Altman, Directeur de Recherche, INRIA Sophia- Antipolis

– Directeur de these : Tijani Chahed, Professeur, Telecom SudParis

Resumé :

The popularity of the great variety of Internet usage brings about a significant
growth of the data traffic in telecommunication networks. Data transmission ef-
ciency will be challenged under the premise of current network capacity and
data fl ow control mechanisms. In addition to increasing financial investment to
expand the network capacity, improving the existing techniques is more ratio-
nal and economical.

Various cutting-edge researches to cope with future network
requirements have emerged, and one of them is network coding. As a nat-
ural extension in coding theory, it allows mixing different network fl ows in the
intermediate nodes, which changes the way of avoiding collisions of data
flows. It is claimed to be a sharp tool which helps achieve better throughput and reliabil-
ity, security, and robustness in various network environments and applications.

This dissertation focuses on the use of network coding for multicast in fixed mesh
networks and distributed storage systems. We first model various multicast rout-
ing strategies within an optimization framework, including tree-based multicast
and network coding; we solve the models with efficient algorithms, and compare
the coding advantage, in terms of throughput gain, in medium size randomly
generated graphs. Based on the numerical analysis obtained from previous ex-
periments, we propose a revised multicast routing framework, called strategic
network coding, which combines standard multicast forwarding and network cod-
ing features in order to obtain the most benefit from network coding at lowest cost
where such costs depend both on the number of nodes performing coding and the
volume of traffic that is coded.

Finally, we investigate a revised transportation problem which is capable of calculating a static routing scheme between servers
and clients in distributed storage systems where we apply coding to support the
storage of contents. We extend the application to a general optimization problem,
named transportation problem with degree constraints, which is widely used
in different industrial fields, including telecommunications, but has not been stud-
ied very often. For this problem, we derive some preliminary theoretical results
and propose a reasonable Lagrangian decomposition approach.