Reinforcement Learning for Slicing in a 5G Flexible RAN

Abstract

Network slicing enables an infrastructure provider (InP) to support heterogeneous 5G services over a common platform (i.e., by creating a customized slice for each service). Once in operation, slices can be dynamically scaled up/down to match the variation of their service requirements. An InP generates revenue by accepting a slice request. If a slice cannot be scaled up when required, an InP has to also pay a penalty (proportional to the level of service degradation). It becomes then crucial for an InP to decide which slice requests should be accepted/rejected in order to increase its net profit. This paper presents a slice admission strategy based on reinforcement learning (RL) in the presence of services with different priorities. The use case considered is a 5G flexible radio access network (RAN), where slices of different mobile service providers are virtualized over the same RAN infrastructure. The proposed policy learns which are the services with the potential to bring high profit (i.e., high revenue with low degradation penalty), and hence should be accepted. The performance of the RL-based admission policy is compared against two deterministic heuristics. Results show that in the considered scenario, the proposed strategy outperforms the benchmark heuristics by at least 55%. Moreover, this paper shows how the policy is able to adapt to different conditions in terms of: (i) slice degradation penalty vs. slice revenue factors, and (ii) proportion of high vs. low priority services.

On using reinforcement learning for network slice admission control in 5G: Offline vs. online

“…Finally, it did not consider the needed physical resources in both UL and DL, which we will introduce in this work. Raza et al 11 proposed a new dynamic SAC model, using reinforcement learning. In this model, the InfProv generates revenues when accepting a slice request and, based on the requested slice priority, pays a penalty when rejecting it.…”

Multi-Tenant Provisioning for Quantum Key Distribution Networks With Heuristics and Reinforcement Learning: A Comparative Study

“…It should be noted that the image used to model secrete-key resources and demands is equivalent to a 2-dimentional matrix. As the image is illustrative to represent resources and demands, this data structure is selected in this study as many recent works addressing other networking issues [57]- [59]. In this work, we assume that multiple TRs arrive dynamically at discrete time steps and can tolerate queuing delay.…”

Control Networks and Smart Grid Teleprotection: Key Aspects, Technologies, Protocols, and Case-Studies

“…5G networks are being designed to support even higher data rates, low latency applications and also connectivity to a miscellany of IoT devices. The new generation of the mobile network will provide three different slices [17]. The first one, named as enhanced Mobile Broadband (eMBB), was released in December 2018 and supports up to 20 Gbps of peak data rate in the downlink and 10 Gbps in the uplink.…”

“…Even with increasing demand, power grids have not matured enough to meet these challenges VOLUME 4, 2016 solutions, covering fast response for mission-critical legacy systems, as well as flexible and elastic connectivity, e.g. Fiber-Wireless (Fi -Wi) networks [13]- [15] and 5th Generation Mobile Networks (5G) [16], [17].…”