Machine learning-based QoS and traffic-aware prediction-assisted dynamic network slicing

Over the last few years, network slicing has been presented as one of the key ingredients in 5G for efficiently specifying network services as per the heterogeneous quality and functional requirements over common shared resources. Network slices are multiple networks having their own management, requirements, and characteristics, positioned over the same physical network with distinct network functions present in each slice. Thus, multiple independent end-to-end networks are supposed to be deployed in 5G, using parallel network slicing. However, it is not easy to guarantee that the traffic on one slice will not affect the traffic on another slice. To implement independent and intelligent network slicing management, this paper proposes a data-driven machine learning-based slicing and allocation model which provides greater flexibility with quality of service (QoS) and traffic-aware reliable dynamic slicing, where resources can be intelligently assigned and redistributed among network slices according to temporal variation of the virtual resource requirements.