Evolution du protocole GTP dans un contexte de VPN et de mobilite-IP

The third Generation UMTS (Universal Mobile Telecommunications System) for mobile communication technology, targets the convergence of telephony, based on an IP paradigm, and also a suite of new services in order to generate new opportunities. In fact, the IMS (Internet Multimedia Sub-System) definition under 3GPP confirms this trend. Moreover, UMTS aims to satisfy the Mobile Network Operators' limitations with an improved spectral efficiency and a lower transport cost, by using the all-IP foundations promoted by Releases 5 and up.;Although the existing UMTS technology is adequate for the traffic types of the 2G and 2.5G, the GTP (GPRS Tunneling Protocol) sub-part being used to support macro mobility with a sustained QoS, has been originally derived from the GPRS (General Packet Radio Service) architecture, and does not fulfill completely the needs of all future Web-Applications. This is due to the fact that it already depends on IP as lower layer, while GTP itself encapsulates IP packets from higher layers. It would thus become crucial to consider an evolution path for GTP, in order to allow a Wireless all-IP Network Access, to be more efficient when based on MPLS (Multi Protocol Label Switching). This would therefore alleviate the current limitations experienced by the GTP protocol mainly during inter-SGSN Hand-Off operations.;The objective of this thesis is mainly to propose and provide a proof-of-concept for an evolution path for GTP, based on MPLS. An algorithm for the establishment of GTP-evolved tunnels promotes the use of a plurality of pre-defined LSP, to be used when needed, especially when inter-SGSN Hand-Offs are triggered by an RAU. Moreover, the intent is to support this new algorithm with an infrastructure that consists of a VPN (Virtual Private Network) arrangement and MIPv6 mobility support, in order to satisfy the triplet of QoS-Security-Mobility during the operation of UMTS Wireless Accesses.;For that purpose, this research is composed of four phases. During the first phase, we will present measurements taken during RAU (Routing Area Updates) at the Gb interface between the BSC/RNC and the SGSN (Service GPRS Support Node). These measurements will serve as benchmark for Hand-Offs. In the second phase, a label-switching model is provided on the Modeler(TM) V10.5 simulation tool from Opnet with the aim of getting a proof-of-concept for static and dynamic pre-defined tunnel set-ups as an evolution path to the classical GTP. In the third phase, we will validate the GTP evolved solution by using the SPIN Model-Checker tool. We may have to consider for that very specific reason, the parallel establishment of tunnels at the time of RAU triggers. Finally, in the last phase, we present an analytical approach for the dimensioning of a voice and videoconference based Wireless-Access Network using a typical router supporting the GGSN (Gateway GPRS Support Node) functions. This approach is based on a concept of effective-bandwidth, which is also integrated in the bandwidth allocation algorithm used for the definition of pre-defined LSP (Label Switched Paths).;The results obtained definitely re-enforce the idea of using pre-defined LSP paths with a better set-up time. Our objective was to perform faster than 1/2 second during an inter SGSN Hand-Off. We are able to switch between paths within less than 0.025 second (more like 1 msec) in a DBBM (Dynamic Break Before Make) mode of operation, and close to 0.0 second in a SMBB (Static Make Before Break) mode of operation. It is however recommended to use DBBM for the Hand-Offs requiring large scalability such as for voice traffic and the SMBB mode of operation for high priority traffic such as 911 calls. The SPIN validation provided us with the requirements for timers and buffers, and also with the need to start the tunnel establishment in parallel with the RAU trigger, in order to avoid conflicting operational states of the context_ack happening before the complete LSP establishment. Finally, the analytical dimensioning demonstrated that the required resources for a population of 1.4 million users in a typical metropolitan area, requires about half of the capacity of a 20 Gbps backplane based GGSN router. Finally, some future framework is presented to support VPN and MIPv6 in a QoS-Security-Mobility capable Wireless-Access-Network.