Design of AI/ML based resource management solutions for beyond-5G networks
No Thumbnail Available
URL
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu |
Master's thesis
Authors
Date
2020-12-15
Department
Major/Subject
Communications Engineering (CE) 2018-2020
Mcode
ELEC3029
Degree programme
CCIS - Master’s Programme in Computer, Communication and Information Sciences (TS2013)
Language
en
Pages
41 + 6
Series
Abstract
Owing to the unprecedented increase in data traffic and the advent of 5G, mobile network operators once again have to overhaul their infrastructure to accommodate the increasing requirements. This expansion requires a significant investment from mobile operators to make their network 5G ready and able to sufficiently serve the growing digital services market. To reduce this cost, one solution is infrastructure sharing in which multiple mobile operators can share common network infrastructure. Sharing real estate side (sites, towers, etc) of the infrastructure between mobile operators has been in practice for quite some time. Sharing network resources is still an immature technique, under research and its implementations differ from case to case. With rapid advances in software defined networking, network slicing is becoming a key tool in infrastructure sharing technology. Using network slicing, separate virtual networks can be created on top of common infrastructure, with each virtual network having its independent resource pool. Two main implementations of network slicing are static and dynamic network slicing. Static network slicing though simplest, does not take varying network requirements into account while dynamic network slicing can dynamically scale the network slice resource as per its requirements. Introducing dynamism adds a layer of complexity in the system. This work introduces a two steps dynamic multi-tenant infrastructure sharing framework. First, it calculates resource allocations for each tenant while monitoring QoS, costs parameters and SLA constraints, second it uses AI to predict the resource sharing ratios among tenants. Experimental evaluations show that results using AI are favourable while being faster than framework using mathematical solver instead of AI.Description
Supervisor
Xiao, YuThesis advisor
Akgül, ÖzgürKeywords
5G, AI, multi-tenant infrastructure sharing, network slicing, resource allocation