Team : Phare
Localisation : Campus Pierre et Marie CurieSorbonne Université - LIP6 Boîte courrier 169 Couloir 25-26, Étage 5, Bureau 516 4 place Jussieu 75252 PARIS CEDEX 05 FRANCE Tel: +33 1 44 27 87 86, Mamadou-Tahirou.Bah (at) nulllip6.fr
Supervision : Thi-Mai-Trang NGUYEN
Definition of a language for the unification of local and cloud networks
The last few years have been marked in the world of networks by the emergence of new architectures based on Software Defined Networking (SDN) and Network Functions Virtualization (NFV). The SDN is characterized by a centralized control plan that is decoupled from the data plan, while the NFV consists of a separation of the network functions from the equipment. These are two different technologies but their effective combination will offer clear prospects in the use of networks. Unlike current networks, also known as legacy networks, the SDN architecture has a centralized control plan. This is one of the major obstacles to its effective deployment. To this, it is also necessary to add the necessary initial investment, which includes the replacement of existing equipment, the acquisition and deployment of new equipment capable of supporting SDN technology. Our work on this thesis is divided into two parts. In the first part, we will explore the different approaches allowing the incremental introduction of the SDN taking into account technical constraints, CAPEX and OPEX costs. This phase is also called the transition phase from legacy to SDN. It implies the simultaneous coexistence of the two technologies within the domain of the same network or next to each other in neighbouring networks, for example. It also involves communication between legacy technologies and the SDN to provide services to users. Hybrid SDN networks are therefore the result of the Coexistence, Communication, and Crossover of these technologies. In the second part of this thesis, we are interested in analyzing the performance of the most advanced SDN controllers in terms of topology management, fault tolerance, throughput, control traffic and scaling. The results obtained indicate that the Open Network Operating System (ONOS) controller is best suited for this transition phase. This is mainly due to the diversity of functionalities such as clustering that it offers but also the performance of the ONOS controller on fault tolerance and scaling.