Maximizing Infrastructure Providers' Revenue Through Network Slicing in 5G

Abstract

Adapting to recent trends in mobile communications towards 5G, infrastructure owners are gradually modifying their systems for supporting the network programmability paradigm and for participating in the slice market (i.e., dynamic leasing of virtual network slices to service providers). Two-fold are the advantages offered by this upgrade: i) enabling next generation services, and ii) allowing new profit opportunities. Many efforts exist already in the field of admission control, resource allocation and pricing for virtualized networks. Most of the 5G-related research efforts focus in technological enhancements for making existing solutions compliant to the strict requirements of next generation networks. On the other hand, the profit opportunities associated to the slice market also need to be reconsidered in order to assess the feasibility of this new business model. Nonetheless, when economic aspects are studied in the literature, technical constraints are generally oversimplified. For this reason, in this work, we propose an admission control mechanism for network slicing that respects 5G timeliness while maximizing network infrastructure providers' revenue, reducing expenditures and providing a fair slice provision to competing service providers. To this aim, we design an admission policy of reduced complexity based on bid selection, we study the optimal strategy in different circumstances (i.e., pool size of available resources, service providers' strategy and trafic load), analyze the performance metrics and compare the proposal against reference approaches. Finally, we explore the case where infrastructure providers lease network slices either on-demand or on a periodic time basis and provide a performance comparison between the two approaches. Our analysis shows that the proposed approach outperforms existing solutions, especially in the case of infrastructures with large pool of resources and under intense trafic conditions.

Peer Reviewed

Postprint (published version)