RT Journal Article
T1 RL-NSB: Reinforcement Learning-based 5G Network Slice Broker
A1 Sciancalepore, Vincenzo
A1 Costa Pérez, Xavier
A1 Banchs Roca, Albert
AB Network slicing is considered one of the main pillars of the upcoming 5G networks. Indeed, the ability to slice a mobile network and tailor each slice to the needs of the corresponding tenant is envisioned as a key enabler for the design of future networks. However, this novel paradigm opens up to new challenges, such as isolation between network slices, the allocation of resources across them, and the admission of resource requests by network slice tenants. In this paper, we address this problem by designing the following building blocks for supporting network slicing: i) traffic and user mobility analysis, ii) a learning and forecasting scheme per slice, iii) optimal admission control decisions based on spatial and traffic information, and iv) a reinforcement process to drive the system towards optimal states. In our framework, namely RL-NSB, infrastructure providers perform admission control considering the service level agreements (SLA) of the different tenants as well as their traffic usage and user distribution, and enhance the overall process by the means of learning and the reinforcement techniques that consider heterogeneous mobility and traffic models among diverse slices. Our results show that by relying on appropriately tuned forecasting schemes, our approach provides very substantial potential gains in terms of system utilization while meeting the tenants' SLAs.
PB IEEE
SN 1063-6692
YR 2019
FD 2019-07-18
LK https://hdl.handle.net/10016/29151
UL https://hdl.handle.net/10016/29151
LA eng
NO The work of V. Sciancalepore and X. Costa-Perez was supported by the European Union H-2020 Project 5G-TRANSFORMER under Grant Agreement 761536. The work of A. Banchs was supported in part by the 5GCity project of the Spanish Ministry of Economy and Competitiveness (TEC2016-76795-C6-3-R)
DS e-Archivo
RD 5 jul. 2024