Proactive ML-Assisted and Quality-Driven Slice Application Service Management to Keep QoE in 5G Mobile Networks
Felipe S. Dantas Silva, Ayuri Bessa, Sérgio N. Silva, Samuel Ferino, Pablo Paiva, Marcos Fernando Machado de Medeiros, Lucas F. M. da Silva, José Melquiades de Rezende Neto, Kevin B. Costa, Charles Santos, Douglas Maciel, Lucileide M. D. da Silva, Anderson Inoue, Roger Immich, Eduardo Henrique da Silva Aranha, Allan De Medeiros Martins, Vicente A. de Sousa, Uirá Kulesza, Marcelo A. C. Fernandes, Marcos Salvador · 2023
Network slicing is a core feature that 3GPP defines in the 5G realm, allowing operators and providers to control traffic resources more granularly. This is achieved by provisioning different network services as independent and isolated network logical partitions atop a shared network physical infrastructure. A network slice must be managed at each constituent part's granularity to ensure an acceptable Quality of Experience (QoE) over time. The state-of-the-art provides plenty of solutions at the network functions and resources management level. In this demo, we propose a service application management solution that applies Machine Learning (ML) to conduct predictive analysis to yield the anticipated detection of service application quality degradation. The solution enables further, fully automatic adjustments with high assertiveness to the network slice structure to maintain QoE. The validation was conducted through test trials on an emulated testbed designed to provide flexibility and dynamics, enforcing network slicing in the Radio Access Network (RAN) and core network tiers and adapting resource allocation policies according to the slice instance's needs.