Projet de recherche doctoral numero :4181

Description

Date depot: 1 janvier 1900
Titre: Vers un Contrôle Programmable pour le Futur Internet et le Cloud Computing
Directrice de thèse: Noemie SIMONI (LTCI (EDMH))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini

Resumé: Context of the study and state of the art The control plane and its associated architecture are key elements for an operator to provide carrier-grade services of high quality, with a trouble-free service integration and timely delivery. 3GPP specifies an architecture called IMS (IP Multimedia Sub-System) as a solution for controlling IP-multimedia services. It is a centralized control plan solution that is separate from the user plane and the application plane. As a result, service integration is easier when it is IMS-based and flexible in transporting the service in the user plane. In particular, the introduction of an Application Service that centralizes service-specific features facilitates service deployment of new services within the existing access networks. IMS is the control solution for Orange group; it is considered as strategic for deploying multi-service control for multi-accesses. At the moment, however IMS is only commercially deployed for fixed voice over IP service. Since its first deployment, the IMS architecture has been recognized as relevant for conversational services, in particular, VoIP services, but it has been more debated for other services like content-based services (which remain to be tested on a large scale). Besides, being the first operator deploying IMS based services commercially often generates extra costs. This issue of cost is regularly raised as one of the constraints that hinder the deployment of non conversational services like RCS (Rich Communication Suite). In this respect mid-term solutions that would reduce the cost of deploying IMS such as virtualization could offer an operator the opportunity to leverage the existing IMS architectures for conversational services. Some academic organizations and operators have started considering virtualization of IMS for such purpose but much more research is required to determine whether this truly provides a viable way forward for the industry. Currently, some research bodies are working towards providing the same functionality as a centralized IMS within a virtualized system or ‘cloud’ in order to optimize/reuse the existing resources with a virtualized system. One of the expected benefits is that such virtualization can be made transparent for the users/customers since the service ‘experience’ remains the same as in a centralized architecture. In practice, ‘virtualising’ a standard IMS architecture must be studied at all levels: - the data base and user profile level (eg HSS), - the control point level (i.e. CSCFs) and/or - the service level (Application Servers) and also to what extent such new mechanisms would be backward-compatible with the current centralized IMS architecture (or even whether such backward-compatibility is desirable). For this, one of the options to investigate is to identify the functionalities in each node and re-arrange them in a different manner in order to mutualize the common functionality and then offer it as one single “interface” to the outside world e.g. the applications. This approach is generally referred to as “infrastructure as a service” is now quite widespread in cloud computing but whether this is truly applicable to the evolution of the IMS control plane as a whole remains to be studied in detail. A number of essential aspects must be considered in making these entities ‘virtual’: the optimized distribution algorithm, the security aspects, the QoS control etc. Recent concepts, which will have to be taken into account in this approach include: - P2P IMS: a model of network architecture which is based on the equivalence of the all devices, including terminals and network elements, so that all the nodes in the network are equal to each other and support the same functionalities. There is no fixed hierarchy or central server. All the elements can act indifferently as a server or a client. The users are identified and the architecture provides mechanisms for locating users or resources within the P2P SIP overlay using the Uniform Resource Identifier (URI) of a user agent. Routing in such architecture is unconventional in the sense that it requires a resolution mechanism to map the identifier with a particular device in real time. - The Advanced Multimedia Systems (AMS): also known as H.325, AMS is the third-generation multimedia system currently under development in ITU-T SG16 (the first generation was H.320 (ISDN), the second generation was H.323 and SIP (VoIP)). H.325 can be considered as a distributed architecture compared to IMS so that the virtualization of IMS architectures may be inspired by the H.325 concept. Although none of these architectures have led to concrete deployments yet, the area of distributed SIP architectures and control plane evolution/virtualization is clearly getting significant interest from the research community. In parallel, the virtualization of control plane can prove very interesting to reduce the control plane cost which

Doctorant.e: Shen Qing