Machine learning based fast self optimized and life cycle management network

• Published in: Computer networks (Amsterdam, Netherlands : 1999) Netherlands : 1999), 2022-05, Vol.209, p.108895, Article 108895
• Main Authors: Nacef, Abdelhakim, Kaci, Abdellah, Aklouf, Youcef, Dutra, Diego Leonel Cadette
• Format: Article
• Language: English
• Subjects: Configurations; Deep learning; Machine learning; Network functions virtualization; Network management systems; Optimization; Service Function Chaining; Software defined networks; Software-defined networking
• ISSN: 1389-1286; 1872-7069
• DOI: 10.1016/j.comnet.2022.108895

6G system targets many emerging industrial verticals, including industry 4.0 and autonomous driving, with extreme service requirements and a high amount of resources, which further overload mobile networks. These industrial verticals are characterized by intense, continuous, and conflicting requirements that harden the mission of the next generation. To achieve the coveted objectives and deal with these shortcomings, the end-to-end network communication should be self-managed, self-orchestrated, and self-optimized, including network edges and clouds. Besides software-defined networking (SDN) and network function virtualization (NFV) that offer network softwarization, elasticity, and flexibility, Machine Learning (ML) is expected to play a vital role in the next-generation of networks. The decisions made at the orchestration plan should be quick besides their accuracy and optimality to overcome the foreseeable challenges. In this paper, we propose a generic framework, named Deep Learning Optimization for Service Function Chaining (DLO-SFC), that provides fast optimal configurations using deep learning techniques and network optimization. We have designed the proposed framework to be generic and orthogonal to a specific use case. Indeed, the framework can be leveraged for any networking configuration and orchestration problems with a slight modification. We evaluate our proposed framework performances showing that it reduces the OPEX costs while optimizing the execution time. [Display omitted] •Efficient AI-based solution for multipaths forwarding in mobile networks.•Machine learning based generic framework for optimal network configuration.•Intelligent Path Re-computation for SDN based networks.