Orchestration of 5G Networks and Beyond

Orchestration of 5G Networks and Beyond

Organized by the ANR MAESTRO-5G project

Program

09:30 – 10:00  Welcome message, virtual coffee & tests

10:00 – 11:00   Keynote 1: George Iosifidis   Delft University of Technology, the Netherlands

Title: Architectures and Algorithms for Software-defined Edge Systems

Abstract: Recent years have seen a striking transformation of the wireless networks landscape. New types of users such as autonomous vehicles and tiny IoT devices are foreseen to create an unprecedented volume of data traffic, while novel bandwidth-hungry and latency-sensitive applications are deployed, literally, every day. In order to address these challenges, the emerging 5G+ systems include several innovations, two of the most promising being: (i) the employment of SDN/NFV technologies to create fully-programmable wireless systems; and (ii) the deployment of services closer to demand by harnessing resources at the network edge. In this talk we will discuss key bottleneck issues arising in this new era, and will present our first theoretical and experimental results for problems related to: (i) virtualization and edge computing in cellular networks, and (ii) mobile IoT analytics in the wild. 

11:00 – 12:15  Session 1

Toward Radio Access Network Slicing Enforcement in 5G System , Imane Oussakel, Philippe Owezarski, Pascal Berthou (LAAS)

Real Traffic Aware Scheduling of Computing Resources in Cloud-RAN , Hatem Kheder, Sahar Hoteit, Patrick Brown, Véronique Vèque (L2S); Ruby Krishnaswamy, William
Diego (Orange); Makhlouf Hadji (SystemX). Slides

Full Coverage hole in Cloud Radio Access Networks , Makhlouf Hadji, Niezi Mharsi (SystemX)

12:15 – 12:30  Invited talk & demo: Toward a Smart and Open Radio Access Network (ORAN) platform, by William Diego (Orange). Slides

The O-RAN Alliance introduces a disaggregated Radio Access Network architecture to drive flexibility, speed and innovation into 5G RAN deployments. To facilitate open implementations of this architecture the O-RAN Alliance and the Linux Foundation have launched the O-RAN Software community focused on aligning a software reference implementation with the O-RAN Alliance’s open architecture and specifications. The aim is to achieve a solution that can be utilized to unify and accelerate the evolution and deployment in the RAN. This talk briefly introduces the O-RAN Alliance, its architecture and key outcomes, as well as showcases its first open source implementation. 

12:30 – 13:30  Lunch Banquet Break

13:30 – 14:15 Technical demonstrations (recorded videos) 

Toward a Smart and Open Radio Access Network (ORAN) platform – Last ORAN release showcase.
William Diego  (Orange). Video

Network Slice Placement using user location – Interactive Gaming Case of Study , José J. Alves Esteves, Amina Boubendir, Fabrice Guillemin (Orange), Pierre Sens (Sorbonne Université). Video

Autonomous Anomaly Detector for Cloud-Radio Access Network on Microservice-based Prototype , Nazih Salhab, Rana Rahim, Rami Langar (Univ. Gustave Eiffel). Video.

14:15 – 15:00   Keynote 2: Roberto Riggio , i2cat, Spain

Title: Splitting the RAN: Functional decomposition in 5G networks       

Abstract: 5G networks are expected to support various applications with diverse requirements in terms of latency, data rates, and traffic volume. Cloud–RAN and densely deployed small cells are two of the tools at disposal of Mobile Network Operators to cope with such challenges. In order to mitigate the fronthaul requirements imposed by the Cloud–RAN architecture, several functional splits, each characterized by a different demarcation point between the centralized and the distributed units, have emerged. However, the selection of the appropriate centralization level (i.e., the functional split) still remains a challenging task, since a number of parameters have to be considered in order to make such a decision. In this keynote, I will discuss the trade-offs associated with functional split selection in small cell networks. Moreover, I will also sketch the current standardization and opensource activities in the field of disaggregated RAN.

Slides

15:00 – 15:50 Session 2

Génération de colonnes pour le problème de routage à délai variable , Nicolas Huin, Jérémie Leguay, Sébastien Martin (Huawei Technologies)

Orchestrez et chainez vos services grâce à votre bon vieux protocole de routage , Adrien Wion, Mathieu Bouet, Vania Conan (Thales), Luigi Iannone (IMT)

15:50 – 16:10  Coffee break

16:10 – 17:00  Lightening talks

Resource Provisioning for Network Slices with Coverage Constraints , Quang-Trung Luu (UPSud, Nokia Bell Labs), Sylvaine Kerboeuf (Nokia Bell Labs), Alexandre Mouradian (UPSud), and Michel Kieffer (UPSud)

Taking into consideration novel requirements in network slice provisioning , Wesley Coelho (Orange/Cnam), Amal Benhamiche, Nancy Perrot (Orange), Stefano Secci (Cnam)

Integrating Non Terrestrial Networks into 5G using Network Slicing , Youssouf Drif, Olivier Gremillet (IRT Saint-Exupery/IRIT); Emmanuel Chaput, Emmanuel Lavinal (IRIT); Pascal Berthou (LAAS); Boris Tiomela Jou (Airbus Defence and Space);  Fabrice Arnal (Thales Alenia Space)

AIDY-F2N: Artificial Intelligence and Dynamic Modeling tools for Future Flexible Networks, Hind Castel, Badii Jouaber, Maxime El Kael (Télécom SudParis); Massinissa Ait Aba (Davidson) 

17:00 – 18:00 Session 3

Function Virtualization Architectures for IoT and Wireless Sensor Networks , Melek Charfi, Alexandre Mouradian, Lynda Zitoune, Véronique Vèque (L2S)

Multi-tenant allocation of virtualized resources for Edge Computing , Andrea Araldo (IMT), Alessandro Di Stefano, Antonella Di Stefano (Univ. di Catania)

Verticals services enabled by customized network slices , Cao-Thanh Phan (IRT b<>com), Philippe Bertin (Orange)

18:00 – Workshop end and virtual cocktail

George Iosifidis is an Assistant Professor with Delft University of Technology, the Netherlands. He obtained his M.Sc. (2007) and Ph.D. (2012) degrees from the Department of Electrical and Computer Engineering, University of Thessaly, Greece, and holds an engineering diploma in avionics from the Greek Air Force Academy. He worked as a Post-doctoral researcher at CERTH/Greece and Yale university for 2 years respectively; as an aircraft engineer from 2001 to 2012; and as an Assistant Professor with Trinity College Dublin, Ireland, from 2016 to 2020. His research interests lie in the area of network optimization and economics with applications to wireless networks and edge computing. His work has appeared in PNAS, Nature Communications and Nature Human Behavior and has received Best Paper Awards in IEEE WiOpt 2013 and IEEE INFOCOM 2017. Dr. Iosifidis is a Guest Editor for the IEEE JSAC Special Issue on Caching, an Editor for IEEE Trans. on Communications and for IEEE/ACM Trans. on Networking, and has received a 2018 SFI Career Development Award.

5G networks are intended to assure the diverse services requirements. For that, RAN slicing is considered as a key enabler for the 5G systems deployment. Even though, the slicing is not a novel concept but its implementation in the RAN still remains chanllenging. In this article, we focus on its enforcement in the 5G context. Taking as requirements the inter-
slice orthogonality, slice satisfaction, scalability empowerment and the inter base station cooperation need for the 5G advanced techniques deployment. We tackle the problem as a 2D bin packing multi-objective optimization. Because of the problem NP- hardness, two heuristics are proposed to fulfill the RAN slicing enforcement requirements. And, good results are obtained based on both heuristics evaluation, thus, enforcing the RAN slicing in 5G system.
Cloud-Radio Access Network (C-RAN) is a promising mobile network architecture that is becoming the foundation of 5G wireless network. It permits to centralize the computing resources in the Baseband Unit (BBU) pool which adds more flexibility and increases network performance. However, as computing resources are now shared among the Radio Remote Heads (RRHs) connected to the BBU pool, efficient scheduling algorithms should be explored in order to meet the deadlines requirements of RRHs’ subframes and to increase network throughput.
In this context, we propose optimal scheduling algorithms for computing resources along with three heuristics. We test the different algorithms as a function of different performance metrics such as the offered throughput, the computing resources occupancy and the number of non-decoded subframes. The evaluation is performed under a real traffic model for the incoming subframes. The obtained results highlight the importance of choosing the appropriate scheduling algorithm and bring recommendations to mobile network operators on the best scheduling algorithm that should be adopted to increase network performance.
In this talk, we address the problem of full coverage hole in Cloud Radio Access Networks.  In this context, and to optimize the network coverage when reducing interferences, network operators need new algorithms that enable to consolidate and re-optimize the antennas radii. This prezentation provides an NP-Hardness complexity proof of the full coverage hole problem and proposes a deep Branch-and-Cut algorithm based on the description of new cutting planes to accelerate the convergence time even for large problem sizes. Simulation results and comparison to the state of the art highlight the efficiency of our approach.

William Diego is currently in charge of coordinating Open Source Networking activities at Orange. He is active in open source projects like CNTT, O-RAN (WG5 Co-chair), ONAP and heading Orange’s open source activities in O-RAN Software Community. He received his PhD degree in Computer Science from IMT Atlantic in 2016 and have worked in the telecommunications industry for over 10 years. He contributed to several European and French research projects like 5G NORMA, 5G EVE and STC, and has also contributed to telecommunication standards (ITU-T, 3GPP, IETF). He authored several scientific publications and patents.

The O-RAN Alliance introduces a disaggregated Radio Access Network architecture to drive flexibility, speed and innovation into 5G RAN deployments. To facilitate open implementations of this architecture the O-RAN Alliance and the Linux Foundation have launched the O-RAN Software community focused on aligning a software reference implementation with the O-RAN Alliance's open architecture and specifications. The aim is to achieve a solution that can be utilized to unify and accelerate the evolution and deployment in the RAN. This talk briefly introduces the O-RAN Alliance, its architecture and key outcomes, as well as showcases its first open source implementation.

William Diego is currently in charge of coordinating Open Source Networking activities at Orange. He is active in open source projects like CNTT, O-RAN (WG5 Co-chair), ONAP and heading Orange’s open source activities in O-RAN Software Community. He received his PhD degree in Computer Science from IMT Atlantic in 2016 and have worked in the telecommunications industry for over 10 years. He contributed to several European and French research projects like 5G NORMA, 5G EVE and STC, and has also contributed to telecommunication standards (ITU-T, 3GPP, IETF). He authored several scientific publications and patents.

Network Slicing can be seen as the use of isolated logical networks sharing Network Slices can be seen as isolated logical networks sharing the same physical infrastructure to provide fully customized services. Using this approach Network Operators will  be able to accurately satisfy a wide range of user's needs in the next generations of networks. However, the efficiency of Network Slicing depends on the management of hosting IT and network resources along with corresponding SLAs and Quality of Service (QoS). We focus on the Network Slice placement optimization problem to give an insight about a latency-aware Network Slice Placement model through a Proof-of-Concept (PoC) illustrated with a time-sensitive use case such Interactive Gaming. Therefore, we showcase our user location-based Network Slice Placement solution compared to the user location-agnostic approach currently used on the state-of-the-art. We show the benefits of considering user location as a primary constraint to ensure QoS and Quality of Experience (QoE).
In this demonstration, we present an autonomous anomaly detector for Cloud Radio Access Network (C-RAN) prototype implemented as microservices using
Docker containers. In addition, we show our in-house developed northbound web application that interfaces to software-defined radio access network controller for
configuration management including network slicing, slices lifecycle management and devices re-homing. Through a real-time updated dashboard used to display network performance indicators, we illustrate metrics processing using an automated data engine. Finally, using conventional off-the-shelf smartphones, we run experiments to present the real-time data ingestion allowing to flag anomalies in time series RAN data caused by external environmental factors.

Roberto Riggio is Senior Researcher at i2cat. Before that he was Head of the Smart Networks and Services Unit at FBK. His research interests revolve around optimization and algorithmic problems in networked and distributed systems. His current fields of applications are edge automation platforms, intelligent networks, and human-driven networking. Within the EU Horizon 2020 5G-CARMEN, he coordinated the first world-wide cross-country validation of 5G for connected, cooperative and automated mobility (CCAM) across the Bologna to Munich 5G Corridor. He has published more than 130 papers in internationally refereed journals and conference proceedings and has generated more than 4 MEuro in competitive funding in the last 5 years. He has received several awards including the IEEE INFOCOM Best Demo Award (2013 and 2019) and the IEEE CNSM Best Paper Award (2015). He serves on the TPC/OC of leading conferences in the networking field and is an associate editor for the Wiley International Journal of Network Management, the Springer Wireless Networks journal, and the IEEE Transactions on Network and Service Management. He is a member of the ACM and a Senior Member of the IEEE.

Avec la promesse d'un réseau plus performant, la 5G va amplifier l'apparition de services avec des besoins plus stricts pour la latence de bout en bout. Satisfaire des contraintes de délais plus strictes nécessite d'intégrer des modèles d'évolution du délai en fonction de la charge plus réalistes. De nombreuses approches permettent de modéliser le délai d'un réseau dans les cas stochastiques ou déterministes, mais peu de travaux ont intégré ces modèles dans l'optimisation globale du routage. Dans cet article, nous proposons une nouvelle méthode de décomposition qui permet d'inclure ces modèles et nous montrons que nous pouvons satisfaire des contraintes plus fortes qu’avec des modèles simplistes.
La gestion des ressources informatiques opérée par les acteurs du cloud, a inspiré les opérateurs de télécommunications dans la conception des réseaux 5G et des futurs réseaux 6G. Tout comme le CPU et le stockage, les ressources réseaux sont de plus en plus mutualisées et attribuées dynamiquement afin de s'adapter à la demande. Ainsi les opérateurs peuvent offrir une personnalisation du réseau à chaque utilisateur (e.g. sous forme de slice) tout en optimisant l'utilisation de leur infrastructure. Afin de réaliser ces objectifs, les techniques développées ont été fortement inspirées de la virtualisation et l'orchestration effectuée dans le cloud.
Ces technologies permettent, certes, la gestion de larges systèmes distribués, mais sont extrêmement dépendantes de la fiabilité des réseaux TCP/IP, sur lesquels ils reposent. Décliner les idées principales de ces méthodes, comme la centralisation du contrôle, sacrifierait la robustesse des réseaux sur l'autel de la performance. Cette fragilisation des services réseaux pourrait donc réduire, par extension, la qualité des services IT qui en dépendent. Dans cette présentation, nous explorerons la possibilité de conserver la robustesse et la stabilité des réseaux existant tout en remplissant les objectifs de performance des réseaux 5G. Dans un premier temps, nous montrerons qu'il est possible de s'appuyer sur les protocoles de routage classiques afin de réaliser du chaînage de service. Notre proposition est à la fois complètement distribuée et interopérable avec les réseaux existants.
Nous comparerons analytiquement les performances de notre solution distribuée avec une approche centralisée. Finalement, nous explorerons la possibilité d'appliquer la théorie du
contrôle afin de piloter dynamiquement l'allocation de ressources réseaux tout en garantissant des propriétés de stabilité.

With network slicing in 5G networks, Mobile Network Operators can create various slices for Service Providers (SPs) to accommodate customized services. Usually, the various Service Function Chains (SFCs) belonging to a slice are deployed on a best-effort basis. Nothing ensures that the Infrastructure Provider (InP) will be able to allocate enough resources to cope with the increasing demands of some SP. Moreover, in many situations, slices have to be deployed over some geographical area: coverage as well as minimum per-user rate constraints have then to be taken into account.This work takes the InP perspective and proposes a slice resource provisioning approach to cope with multiple slice demands in terms of computing, storage, coverage, and rate constraints. The resource requirements of the various SFCs within a slice are aggregated within a graph of Slice Resource Demands (SRD). Infrastructure nodes and links have then to be provisioned so as to satisfy all SRDs.Two provisioning approaches are considered and compared to a baseline approach. The sequential approach provisions resources slice by slice and the joint approach considers the constraints of all slices simultaneously. Numerical results show the effectiveness of the proposed provisioning framework for a slice deployment on a mobile network infrastructure satisfying a minimum data rate for users in the geographical areas where services have to be made available.

In this talk we will discuss novel mapping and provisioning requirements arising with new 5G radio and core policies. By addressing the related challenges as an optimization problem, we present an approach encompassing flexible functional splitting, network function decomposition and control-plane and data-plane separation. We then demonstrate with numerical simulations the impact of taking into full and partial consideration the new possibilities of network slice design.

With the imminent deployment of the 5th generation mobile network (5G) in the non-standalone version, some researches focus on network slicing to fully exploit the 5G infrastructure and achieve a higher level of flexibility in the network. The release 17 of 3GPP integrates Non Terrestrial Networks (NTN) in 5G networks and more specifically satellite networks. In this context, we have worked on the network slicing concept applied to satellite networks in order to provide a seamless integration of the satellite network into 5G networks as a transport network. In our work, we  propose a novel Satellite Slicing Framework in order to fully exploit the satellite infrastructure and to facilitate the integration of satellite services into 5G. We enumerate key challenges related to network slicing applied to satellite networks, propose solutions or guidelines to address those challenges and ultimately implement a proof of concept to demonstrate an end-to-end 5G slice with a satellite network as a transport network.

In this presentation, we tackle the problem of both IoT and Wireless Sensor Network architectures by defining the ``networking functions'' necessary in order to support the diversity of IoT applications in the context of smart cities. We also survey some prominent network architectures and analyze how and where the defined networking functions fit in these propositions, and we finally propose an alternative approach using the Software-Defined Networking paradigm to implement these functions.
Edge Computing (EC) consists in deploying computation resources at the network edge (base stations, access points, central offices), in order to run network functions or applications closer to the users. The aim is to reduce upstream traffic and increase Quality of Experience (QoE). To enable EC, resources at the edge are virtualized, e.g., using containerization, and are made available to functions and applications. However, resources are limited at the edge, and the problem arises on devising strategies to allocate them to different functions/applications. We present the recent advances in the literature related to this problem and the various approaches adopted (game theory, linear programming, stochastic
optimization).
We finally present on our recent contribution, which focuses on a Multi-tenant setting, in which a Network Operator (NO) owns physical resources, virtualizes them and let 3rd party Application Providers (APs) run their applications on top of them. The NO decides the allocation in order to maximize its own utility, i.e., traffic saving, and APs adapt to the resources granted in order to maximize their own utility, i.e., QoE for their users. We devise strategies for the NO to optimize the allocation. The resource allocation problem in this setting is complicated by the fact that we adopt the realistic assumption that APs are not willing to let the NO know the data and the computation they are running at the edge. Therefore, NO must be able to allocate resources with no direct visibility of how they are used.  We compare the different methods that we adopted for this problem,resorting to measurement-driven stochastic optimization, heuristic algorithms
and linear programming.
In this presentation we introduce the 5G dynamic system designed in 5G-TRANSFORMER project to be able to offer network slices tailored to the specific needs of various vertical industries. The technical approach relies on innovations around 3 components: 5GT-VS (Vertical Slicer) offering a powerful, yet simple and flexible, interface to verticals;  5GT-SO (Service Orchestrator) capable of instantiating and orchestrating network services, including federation mechanisms; 5GT-MTP (Mobile Transport and computing Platform) integrating compute, storage and networking resources. After detailing the general architecture we show its  benefits to address different vertical use cases, namely ehealth, entertainment, automotive and MVNO. Finally we present the techno-economic study showing network slices cost breakdown and its evolution over time.