Skip to Main content Skip to Navigation
Theses

Contrôle de congestion et gestion du trafic à partir de mesures pour l'optimisation de la QdS dans l'internet

Abstract : Internet measurement has only been used for research, engineering and design of Internet networks since few years (since the beginning of years 2000), but it is more and more popular and spreads rapidly. It deals with studying, characterizing, analyzing and modeling tra_c on the di_erent Internet links in order to understand network behaviors when facing tra_cs which are largely unknown at this time. In particular, guarantying QoS (Quality of Service) for the Internet is currently one of the most challenging issues. Internet tra_c monitoring and analysis show that current transport mechanisms (TCP) introduce Long Range Dependence (LRD) properties which induce very high tra_c variability and force to largely over-provision communication resources [20]. This thesis aims at designing new communication protocols and architectures able to reduce the tra_c LRD in order to optimize the use of communication resources. Then, new protocol and architectural mechanisms could be perfectly suited to users' needs and tra_c constraints [Lar03a]. Finally, this is the research and network engineering process which will be modi_ed by integrating a new initial monitoring stage for collecting data and some knowledge on existing networks and tra_c ; it will help, in a second time, to design and deploy new optimized networks. Thus, this PhD work deals with a new approach for the Internet, aiming at improving tra_c management, QoS and more generally network services. This approach, called Measurement Based Networking (MBN), is built on the use of active and passive monitoring techniques to evaluate in real time di_erent network parameters and analyze its tra_c in order to react very quickly and accurately to speci_c events arising in the network (for instance, congestion events). We will illustrate, in particular, the MBN approach by designing a new measurement based congestion control mechanism (MBCC) which will be evaluated thanks to NS-2 simulations [Lar05c].We will show, in particular, how this new mechanism can improve tra_c characteristics as well as Internet QoS, despite the complexity and variability of current Internet tra_cs. Finally, in the last part of this PhD work, we will present how the MBN approach and MBCC can guaranty a robust QoS, i.e. are able to provide the requested QoS in all circumstances, including in the presence of Denial of Service (DoS) attacks [Owe04f]. Indeed, we will use the MBA architecture, based on real time tra_c measurements, to adapt to normal or abnormal disruptions which arise in the network. In particular, we will demonstrate that the MBCC congestion mechanism, designed for generating regular and optimized tra_c, makes the Internet more robust than TCP when facing DoS attacks.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00009745
Contributor : Emilie Marchand <>
Submitted on : Wednesday, July 13, 2005 - 10:46:20 AM
Last modification on : Friday, January 10, 2020 - 9:08:09 PM
Long-term archiving on: : Friday, April 2, 2010 - 10:16:33 PM

Identifiers

  • HAL Id : tel-00009745, version 1

Citation

Nicolas Larrieu. Contrôle de congestion et gestion du trafic à partir de mesures pour l'optimisation de la QdS dans l'internet. Réseaux et télécommunications [cs.NI]. INSA de Toulouse, 2005. Français. ⟨tel-00009745⟩

Share

Metrics

Record views

428

Files downloads

8937