Analyzing the PMachine-Learning-Based 5G Network Function Scaling via Black- and White-Box KPIsower Consumption in Cloud-Native 5/6G Ecosystems

The diffusion of the Fifth-Generation (5G) of
mobile radio networks will be the main driver in the digital
transformation towards a new hyper-connected society. In
order to satisfy the stringent demands of 5G-ready applications
over the limited resources available at the edge, scaling
mechanisms become crucial to guarantee the performance levels
envisaged for 5G. Such mechanisms must be able to
automatically perform according to the real-time user demands,
the availability of computing resources and the state of Network
Functions (NFs) and applications. In this context, this paper
proposes a deep learning model, based on Artificial Neural
Networks (ANNs), for the dynamic and automated
orchestration of NFs. The novelty of this model is its
independence from specific 5G NF implementations; this is due
to the nature of the Key Performance Indicators (KPIs) used in
this work, which are related to both execution environment
(standard “black-box” KPIs) and standard 5G APIs (“whitebox”
KPIs). Results obtained on the orchestration of a Session
Management Function (SMF) reach an accuracy of 97~98% for
the training and validation phases and above 95% for the
deployed model, as well as higher overall accuracy by ~5% and
computational resource savings with respect to a thresholdbased
scheme.