A Softwarized Service Infrastructure for the Dynamic Orchestration of IT Resources in 5G Deployments

Abstract

Thanks to the 5G, telco operators can offer a new set of advanced services to mobile users which makes use of heterogeneous IT resources deployed at the edge of the network. However, their optimal management is not a simple task to accomplish because of the extreme variability characterizing offered services, traffic profile, user distributions, bandwidth, computing, and memory capabilities available for nodes hosting IT resources. To provide preliminary answers in this direction, this paper presents a high-level description of a softwarized service infrastructure, based on the ETSI-NFV specifications, able to dynamically orchestrate IT resources. Specifically, the number of users attached to the base stations and the capabilities of nodes hosting IT resources are continuously monitored through Software-Defined Networking facilities and reported to a high-level orchestrator. Here, a Convolutional Long Short-Term Memory scheme is firstly adopted to provide a spatio-temporal prediction of user distributions and related traffic demands. Then, an optimization problem is executed for configuring location, settings, amount, and usage of IT resources, based on the prediction outcomes. The behavior of the prediction process is deeply investigated. The optimization problem, instead, is described in its preliminary formulation, which gives a clear idea of future research activities in this direction.