Tutorial 2

Vertical-Oriented End-to-End Orchestration in 5G Networks: Modeling, Optimization, Implementation, and Verification

Monday, 18 June 2018, 14:00-18:00, E4 hall

Speakers:

• Vincenzo Sciancalepore (NEC Laboratories Europe, Germamy)
• Mérouane Debbah (Huawei Technologies, France)
• Alessio Zappone (CNRS, France)
• Marco Di Renzo (CNRS, France)

Motivation and Context

The massive deployment of “smart” devices, broadband and mission critical services along with a huge variety of scenarios, ranging from smart city to broadband media are paving the way for a novel and disruptive 5G communication network that will enable massive capacity, zero delay, faster service development, flexibility, elasticity and optimal deployment, less energy consumption, enhanced security, privacy by design, and connectivity to billions of devices with less predictable traffic patterns. Accordingly, next-generation networks need to be capable of handling a complex context of operations and support an increasingly diverse set of new and yet unforeseen services, whose extremely diverging requirements will significantly boost mobile network performance and capabilities.
Additionally, next-generation networks need to provide flexible, smart and scalable adaptation and/or association of the available network resources to the specific requirements of the supported services, enabling a dramatic paradigm shift from the CAPEX to the OPEX “Everything-as a-Service” driven business models
To meet the 5G challenges, the ICT industry has moved fast, developing a number of breakthrough innovations. Multiple vendors from all over the world bring the new technology into the market even though most of the products are still in the development phase, whereas a global consensus is progressing slowly on how 5G will proceed with respect to technological barriers, regulatory restrictions and standardization activities. It is clear that 5G networks will not be based on a single specific technology, but they will be considered as a portfolio of access, connectivity and flexibility solutions addressing the demands and requirements of mobile communications beyond 2020. In order to cover the requirements from different application domains, technologies promoted in the 5G landscape range from advances on the radio access network, such as ultra-lean radio access design, device-to-device communications, MIMO antennas, along with the benefits of slicing management.
In parallel to 5G networks evolution, there are application verticals that demand specific and contradicting characteristics. According to the 5G Manifesto, 11 industry segments are considered as first-class citizens for early experimentation and large-scale demonstration for European industry and society. However, different verticals have different levels of readiness and interest for 5G adoption. With the proliferation of smart devices, the Media vertical transforms and more users are switching from the traditional linear broadcasting services (TV channels) to OTT (Over The Top) streaming services, provided by traditional broadcasting companies (e.g., Catch-up TV services) or by global companies (i.e., Apple, Amazon, Google, Netflix, etc.); the video quality increases (i.e. 3D, UHD/4K/8K/12K, Augmented and Virtual Reality (VR), etc. ), which can be translated to increased bandwidth requirements for both the core and access networks.
Finally, cloud services allow each individual user to easily create video content, store it and share it through various social network platforms. Technological advances, political visions and market liberation are transforming the Energy vertical from a closed, monolithic and highly predictable infrastructure to an open, multi-owned, decentralized ecosystem, able to support Smart Grid mission critical applications (such as fault localization, isolation/self-healing and energy re-routing), massive IoT smart metering applications requesting more stringent capacity and privacy, along with smart Electrical Vehicles (EV) charging, posing huge challenges, both in functional (i.e. stability, resiliency and highly availability) and in non-functional (i.e. security, privacy and CAPEX/OPEX) directions.
This emerging ecosystem of vertical-oriented design of 5G networks is little understood in academia and industry, and, to the best of our knowledge, there is a fundamental lack of approaches for modeling and optimizing it with focus on the diverse requirements of different vertical markets. This tutorial aims at filling this gap, describing in detail the enabling techniques to allow multiple vertical-oriented services to coexist by sharing the same physical infrastructure.

Structure and Content

The tutorial is organized in four main parts that are described as follows.
Introduction. The tutorial will provide a brief introduction of the problem of vertical-oriented orchestration design of 5G networks, motivating its relevance for future wireless communications. 5G networks are expected to provide great support to the operation and management of 5G end-to-end heterogeneous facilities enabling advanced 5G services development, easier, safer and more secure testing and verification from an operational and financial point of view. In addition, the orchestration solution will orchestrate different heterogeneous 5G network facilities able to provide highly available services, to support multi-tenancy, to manage homogeneously virtualized multi-homed, static or moving, hardware constrained (smart energy, media and transport) devices, to integrate multiple RAN technologies, to manage edge computing resources and to manage virtualized services in an elastic way for the needs of media, smart city and energy vertical scenarios.
This part of the tutorial will describe the virtualization issues of Smart Energy, Virtual & Augment Reality (VAR) and Smart City use cases, covering Enhanced (or Extreme) Mobile Broadband (eMBB), Ultra-reliable and Low Latency Communications (URLLC) and Massive Machine Type Communications (mMTC) as defined by 5GPPP and ITU. These evolving networking requirements raise critical challenges that urge for 5G physical networks (ranging from RAN to Core Network and slicing requirements [1]) along with 5G softwarization, virtualization and Management and Network Orchestration (MANO) requirements, which in turn, can be translated as facilities requirements for i) multi-tenancy that considers the ability to be open, avoid vendors’ lock-in strategies, combine resources from different operators and offer connectivity services to multiple tenants, ii) flexibility, multi-domain and easy-to-use nature in order to dynamically configure the infrastructure in time and space to accommodate innovative foreseen and unforeseen 5G vertical services, iii) high availability to support via network slicing mission critical applications in the area of energy and city transportation.
Network Virtualization and Programmability. Network virtualization has led to significant benefits in terms of business support, service innovation, infrastructure and operational cost reduction but at the same time, it has brought significant technical challenges. The network virtualization paradigm revolutionizes networking and service experience, by abstracting and separating logical network behaviors from the underlying physical network resources. This significantly impacts on the reduction of the capital and operations expenditures, while also driving programmability and automation of network facilities in order to enable a broad range of business services. Network virtualization elevates the traditional monolithic network design into a flexible “network of functions” following the virtual network function paradigm raising several technical challenges.
In this part of the tutorial, we will focus on the pivotal aspect of the network virtualization, i.e., the newly evolved concept of network slicing [2] that paves the road towards an automatic and flexible solution in charge of allocating a specific amount of isolated and/or shared network resources, tailored for particular vertical-oriented service requirements [3]. Network resources are meant as computing and storage capacity, virtualized network functions, physical radio resources, core network functions and backhaul/backbone connectivity. The technical challenges also stretch on the entire lifecycle process, including network slice instantiation and maintenance [4], orchestration and allocation of shared and isolated resources [5], including also communication interfaces amongst different network slices. Finally, we will analyse the implementation details of first network slicing proof-of-concept (PoC) by shedding the light on advantages and drawbacks of implementing such a novel concept on 3GPP-compliant equipment [6].
Network Optimization. This part of the tutorial addresses the issue of network optimization. Specifically, the tutorial will show how to design the system radio resources in order to optimize key performance indicators of 5G networks, including spectral efficiency, energy efficiency, end-to-end latency, etc, while at the same time guaranteeing the desired quality of service of different vertical services sharing the same physical resources, which in turn translates into the desired quality-of-experience for the network end-users. The resulting problems are highly non-convex and require specific optimization approaches beyond the traditional convex optimization framework. We will introduce the frameworks of multiobjective optimization, sequential optimization, fractional optimization, monotonic optimization, which together enable the best trade-off between global performance and computational complexity, while at the same time guaranteeing the individual performance of co-existing vertical sectors. In particular,
-Multi-objective optimization enables flexible trade-offs between individual (at a user-level and/or tenant-level) service requirements [7].
-Sequential optimization has been recently shown to achieve near-optimal performance in interference-limited networks, while requiring affordable complexity [8].
-Fractional programming has emerged as the most suited tool for the optimization of the bit-per-Joule energy efficiency of a wireless network [9].
-Monotonic optimization enables the development of lower-complexity off-line globally-optimal solution to benchmark the performance of online optimization tools [10].
This part of the tutorial provides the audience with a solid background on the above-mentioned optimization frameworks, enabling the audience to formulate and solve practical optimization problems for vertical-oriented 5G network design. Both centralized and distributed designs will be developed, discussing the resulting complexity-performance trade-off.
Network Modeling. To quantify the potential gains of vertical-oriented network modeling and design a critical role is played by the characterization of the spatial distribution of user and infrastructure locations in a network, which critically affects the quality-of-service and quality-of-experience that can be guaranteed to each user. In this context, the theory of point processes and stochastic geometry constitute an essential mathematical tool, as it provides us with general and accurate methods for modeling random spatial patterns.
This part of the tutorial provides the audience with a solid background and comprehensive description of stochastic geometry modeling, by introducing key theorems, by explaining how to formulate problems from the standpoint of system-level analysis and optimization, with major focus on illustrating how to use stochastic geometry to model vertical-oriented 5G networks. In addition to the basic theory, several new methodologies for system-level modeling are illustrated, which include the equivalent-in-distribution approach for error probability computation [11], the moment generating function for area spectral efficiency computation [12], and the intensity-matching approach for quantifying the trade-offs in terms of achievable rate and harvested energy [13]. As far as system-level optimization is concerned, a new definition of coverage probability and spectral efficiency will be introduced and shown with several examples to constitute a tractable approach for the optimization of vertical-oriented large-scale wireless networks [14]. Finally, the suitability of stochastic geometry for wireless networks modeling is substantiated with the aid of experimental data for the locations of cellular base station and for the footprints of buildings, which are taken from two publicly available databases from the UK (OFCOM and Ordnance Survey) [15]. With the aid of several examples, we show how the proposed approach constitutes a holistic framework towards the design of vertical-oriented 5G wireless networks.
The tutorial ends with an open discussion of the latest research directions and open issues that in our opinion represent the most important challenges that need to be addressed and solved, aiming towards the successful design of vertical-oriented 5G networks.