Toward Elastic Distributed SDN/NFV Controller for 5G Mobile Cloud Management Systems

Abstract

Future 5G mobile network architecture is expected to offer capacities to accommodate the inexorable rise in mobile data traffic and to meet further stringent latency and reliability requirements to support diverse high data rate applications and services. Mobile cloud computing (MCC) in 5G has emerged as a key paradigm, promising to augment the capability of mobile devices through provisioning of computational resources on demand, and enabling resource-constrained mobile devices to offload their processing and storage requirements to the cloud infrastructure. Follow-me cloud (FMC), in turn, has emerged as a concept that allows seamless migration of services according to the corresponding users mobility. Meanwhile, software-defined networking (SDN) is a new paradigm that permits the decoupling of the control and data planes of traditional networks and provides programmability and flexibility, allowing the network to dynamically adapt to change traffic patterns and user demands. While the SDN implementations are gaining momentum, the control plane is still suffering from scalability and performance concerns for a very large network. In this paper, we address these scalability and performance issues in the context of 5G mobile networks by introducing a novel SDN/OpenFlow-based architecture and control plane framework tailored for MCC-based systems and more specifically for FMC-based systems where mobile nodes and network services are subject to constraints of movements and migrations. Contrary to a centralized approach with a single SDN controller, our approach permits the distribution of the SDN/OpenFlow control plane on a two-level hierarchical architecture: a first level with a Global FMC Controller (G-FMCC), and a second level with several Local FMC Controllers (L-FMCCs). Thanks to our control plane framework and Network Function Virtualization (NFV) concept, the L-FMCCs are deployed on-demand, where and when needed, depending on the global system load. Results obtained via analysis show that our solution ensures more efficient management of control plane, performance maintaining, and network resources preservation.