Management and Orchestration of Network Slicing in Public-Private 5G Networks

Abstract

5G era is touted as the generation of mobile networks that will deliver an end-to-end ecosystem to enable a fully mobile and connected society. One of the most noticeable differences with respect previous generations is the irruption of the vertical industries (e.g., industry 4.0, transportation, energy, health) in the telco landscape. These verticals aspire to embrace 5G capabilities to shape and accelerate their digital transformation, in their mission to move towards a more modern business ecosystem, grounded upon service innovation and sustainability principles. The consequences of this paradigm shift are clear: in addition to supporting the evolution of the established prominent mobile broadband use cases, 5G will need to support countless vertical use cases with a high variety of applications and variability of their performance attributes, ranging from low bitrate high latency services to high bitrate low latency services, with many variants in between. These use cases will be also delivered across a wide range of devices (e.g., smartphone, IoT sensors, industrial equipment, vehicles), with quite different mobility patterns and energy consumption requirements each. Furthermore, 5G use cases will be provisioned using distributed infrastructures, including carrier networks but also 3rd party nodes that span beyond operators’ footprint. With this scene in mind, the challenge that lies ahead is how to accommodate all this casuistry in 5G, when most of these use cases will be active at the same time. The “onesize- fits-all” architectural approach that exists in today’s networks is not feasible. The reason is that unlike 4G, which is user-centric, 5G will be industry-centric, so the design of one single physical network optimized to process mobile broadband traffic no longer makes sense. Additionally, the vertical use cases will have quite different (sometimes conflicting) service requirements, tied to Service Level Agreements (SLAs) that are much less flexible that those existing in traditional mass-market services. One solution could be to design dedicated networks per use case, following solutions such as those existing with DECOR. However, this solution is neither scalable (more and more use cases are defined every year) nor viable (too much upfront costs) for operators. In such a case, fresh solutions need to be explored. This is where the concept of network slicing fits in. Network slicing is a solution whereby a physical network infrastructure is split into a set of logical network partitions, each tailored to satisfy the specific service requirements of a given vertical or use case. These partitions, referred to as network slices, are potentially operated isolated from each other but instantiated and running over the same physical network. ¡In 5G, the network will be a continuum spanning across different administrative domains. These domains typically correspond to infrastructures managed by the different mobile network operators, referred to as public land mobile networks (PLMNs); however, some of these domains can also correspond to private infrastructures managed by the verticals themselves, such as factories and transportation hubs (seaport, airport, etc.). To make slicing happen, in thus needed to build an end-to-end (E2E) open infrastructure that integrates networking, computing and storage resources, together with technologies for their segregation and programmability, to transform networks into a flexible, reliable and secure well-orchestrated facility across multiple administrative domains. Within this high-level objective, the goal of this dissertation is to design and validate solutions for network slicing management and orchestration in multi-domain environments, with applicability in public and private 5G network scenarios. To that end, this thesis is structured into three main workstreams. The first one corresponds to the design of system architecture solution or multi-domain network slicing. This architecture will build upon network softwarization technologies (Software Defined Networking and Network Functions Virtualization) together with their orchestration-enabling artifacts, and integrate them into a robust, scalable, and standardscompliant system. The resulting system will provide all the necessary capabilities that shape network slicing concept (including isolation, customization, elasticity, programmability, and automation, among others), with multiple service-tailored logical networks running atop a distributed yet common physical network. Special focus is put on the aspects related to the lifecycle management of these network slices, from their design to their provisioning until their termination, with operation in between. The second line of work will be focused on the implementation of the system architecture, and the prototype validation in 5G experimentation facility. The selected environment is 5G-VINNI. 5G-VINNI is a large-scale E2E infrastructure consisting of 5G nodes that are distributed across Europe, and that provides a testing and validation environment for vertical use case experimentation. For the system implementation, Open Source MANO (OSM) stack will be combined with Openslice. The result is an E2E management solution suite providing the network and service orchestration capabilities which are needed to partition infrastructure resources and allocate them to different slices, at both provisioning and operation time. This solution will incorporate federation capabilities, to facilitate operation in multi-domain environments. Finally, a Proof-of- Concept (PoC) will be set up, to showcase the behavior of the solution in different use cases: network slice design (creating a network slice descriptor and onboarding it to the catalog), network slice provisioning (commissioning a network slice upon a verticaltriggered service order) and scaling (increase the capacity of the slice, by allocating more resources). The PoC also exhibits also multi-domain aspects; to that end, Spanish and Greek nodes from 5G-VINNI facility have been selected. The third and last line of work focuses on the study of the private 5G market, and the analysis of network slicing role in it. Unlike private Long-Term Evolution (LTE), based on the use of infrastructures totally separated from the public LTE network, the expected continuum in 5G will span nodes from public and private infrastructures. Actually, there are many verticals that require executing E2E services spanning these two types of infrastructures, with some workloads running on-premises (private infrastructure) and some others on the operator’s footprint (public infrastructure). To break silos in this publicprivate network infrastructures and facilitate a seamless provisioning of E2E service across them, slicing is identified as “the solution”. However, for this to happen, we need to first understand the private 5G ecosystem, why verticals prefer having standalone private 5G networks in the short term, and their motivations to partially migrate 5G public network in the medium and long term. Based on this understanding, which includes findings from business and technology viewpoints, we will outline a radar for network slicing roll-out in telco networks, to accompany the verticals in the transitions, managing their expectations of what capabilities will be available by when, and under which conditions.