User-Based Quality of Service Aware Multi-Cell Radio Access Network Slicing

• Published in: IEEE eTransactions on network and service management Vol. 19; no. 1; pp. 756 - 768
• Main Authors: Papa, Arled, Jano, Alba, Ayvasik, Serkut, Ayan, Onur, Gursu, H. Murat, Kellerer, Wolfgang
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.03.2022; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Network slicing; Optimization; Quality of service; Quality of service architectures; RAN slicing; Resource management; SD-RAN; Signal to noise ratio; slice isolation; Throughput; Time-frequency analysis; User requirements; User satisfaction; Wireless communication
• ISSN: 1932-4537; 1932-4537
• DOI: 10.1109/TNSM.2021.3122230

5G radio access network (RAN) slicing envisions a solution to flexibly deploy heterogeneous services as slices sharing the same infrastructure. However, this level of flexibility renders slice isolation challenging, mainly due to the stochastic nature of wireless resources. In the state-of-the-art, RAN slicing algorithm's efficiency with respect to slice isolation is related to the ability of meeting individual slice requirements. However, mostly an aggregated slice performance guarantee is considered instead of per user guarantees. Hence, state-of-the-art approaches might not always provide the satisfaction of all users within a slice. Indeed, our results demonstrate that if user requirements within a slice are not included in the RAN slicing algorithm, the per user quality-of-service (QoS) may not be fulfilled. In this paper, we investigate the definition of slice isolation as the ability to satisfy individual users' throughput within slices, in a frequency selective, multi-cell wireless scenario with focus on maximizing slices' throughput. Our problem is tackled with a Lyapunov optimization approach, which proves to always achieve slice isolation. Our results show that our solution does not only achieve 100% user QoS guarantees compared to 50% achieved in the state-of-the-art, but also doubles the throughput with increasing number of BSs.