WeC3- Orchestration and Virtualisation

Wednesday, 19 June 2019, 14:00-15:30, Room 3


Session Chair: Toktam Mahmoodi (King’s College London, United Kingdom (Great Britain))


Network Slicing-aware NFV Orchestration for 5G Service Platforms

Hamzeh Khalili, Apostolos Papageorgiou and Muhammad Shuaib Siddiqui (Fundació i2CAT, Internet i Innovació Digital a Catalunya, Spain); Carlos Colman Meixner (University of Bristol, United Kingdom (Great Britain)); Gino Carrozzo (Nextworks, Italy); Reza Nejabati and Dimitra Simeonidou (University of Bristol, United Kingdom (Great Britain))
The advent of 5G promises to unleash highly pervasive network coverage and density, increased data rate and capacity, optimized instantiation of virtualized resources in a multi-tenant and multi-service network capable of fulfilling the stringent requirements of various heterogeneous vertical applications. Network slicing is widely recognized as one of the key enablers of such 5G to allow multiple isolated virtual networks to be created on the a single shared physical network infrastructure. In this work, we present a survey of various approaches for slicing, we review the management and orchestration tools available to implement network slicing, and describe the Network Slicing architecture designed for the 5GCity project. Our network slicing framework embraces creation of slices of various network elements such as compute nodes, physical networks, radio parts and network edge resources. The slicing concept described in this paper is being evaluated in three live city pilots (Barcelona-ES, Lucca-IT, Bristol-UK) which represent the trial environment for the 5G use cases of the 5GCity project.

5G-TRANSFORMER Service Orchestrator: Design Implementation and Evaluation

Josep Mangues-Bafalluy (Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Spain); Jorge Baranda (Centre Tecnològic de Telecomunicacions de Catalunya (CTTC/CERCA), Spain); Iñaki Pascual (CTTC, Spain); Ricardo Martinez (Centre Tecnològic de Telecomunicacions de Catalunya (CTTC/CERCA), Spain); Luca Vettori (Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Spain); Arturo Zurita and David Salama (ATOS, Spain); Dmitriy Andrushko (Kharkiv National University of Radio Electronics, Ukraine); Konstantin Tomakh (Mirantis, Ukraine); Giada Landi (Nextworks, Italy); Kiril Antevski (Universidad Carlos III, Spain); Jorge Martín-Pérez (Universidad Carlos III de Madrid, Spain); Barbara Martini (CNIT, Italy); Xi Li (NEC, Germany); Josep Xavier Salvat (NEC Labs Europe, Germany)
5G networks will pose complex network management challenges due to the variety of vertical services they will need to serve and the diversity of underlying infrastructure. The service orchestration functionality is fundamental to enable fulfilling the requirements of the former while efficiently using the infrastructure resources. This paper details the 5G-TRANSFORMER service orchestrator implementation and operation. It also evaluates and profiles service creation time showing how the automation offered by the platform allows reducing it from hours to minutes. It also shows that the most time-consuming steps correspond to the deployment of the virtual network functions and post-deployment configuration, which consume one order of magnitude more time than the rest of steps (e.g., network creation, port creation).

Inter-Business Orchestration for Resource and Service Provisioning in 5G Network Slicing

Enrique Chirivella-Perez (University of the West of Scotland, United Kingdom (Great Britain)); Pablo Salva Garcia (University West Of Scotland, United Kingdom (Great Britain)); Jose Alcaraz-Calero and Qi Wang (University of the West of Scotland, United Kingdom (Great Britain)); Pedro Neves (Altice Labs, Portugal)
The Fifth Generation (5G) networks leverage softwarisation and virtualisation as cornerstones to build cost-effective and flexible infrastructures and services. Meanwhile, advanced orchestrator is entailed to manage and orchestrate the resources and services at various levels of infrastructures, including the physical layer, virtualisation layer, service layer etc. The challenges become more evident when the multiple business roles, enabled by the various levels of operations in this new business model paradigm, are taken into account in an integrated operational environment. This paper proposes a new layered, inter-business vertical orchestration architecture to address the highlighted challenges in softwarised and virtualised 5G networks. The architecture distinguishes not only the infrastructure levels but also the business roles, both in an integrated framework with interactions among them specified. It is noted that the proposed vertical orchestration architecture is compatible with network slice orchestration in 5G networks. The proposed orchestration framework is prototyped in a testbed with realistic 5G infrastructures, and its performance is empirically evaluated focusing on the service creation time 5G Key Performance Indicator (KPI).

NGPaaS Framework for Enriched and Customized Virtual Network Functions-as-a-Service

Paul Veitch (BT, United Kingdom (Great Britain)); Adam Broadbent (British Telecom, United Kingdom (Great Britain)); Angelos Mimidis and Jose Soler (Technical University of Denmark, Denmark); Marco Mobilio (University of Milano-Bicocca, Italy); Alessandro Tundro (Unimib, Italy); Michell Guzman (Unimib, United Kingdom (Great Britain)); Steven Van Rossem (Ghent University & iMinds – IBCN, Belgium)
This paper describes how the novel Next Generation Platform-as-a-Service (NGPaaS) framework can facilitate major benefits for Network Operators and Vertical Service Providers (VSPs) who wish to leverage Virtual Network Functions-as-a-Service (VNFaaS) capabilities. Network Operators can benefit by providing an on-demand PaaS with required features for the VSPs, thus generating new revenue streams but with low operational overhead due to the high degree of automation. VSPs can benefit from the PaaS-oriented approach, by being able to flexibly on-board new VNF types and “value-added” service capabilities like monitoring, healing and profiling, to deliver customized service blueprints to meet the needs of their end customers. The paper outlines the design of an early prototype, built on the Open-CORD platform and using industry-standard Virtualised Network Functions (VNFs).

vFPGAmanager: A Hardware-Software Framework for Optimal FPGA Resources Exploitation in Network Function Virtualization

Spyros Chiotakis (Virtual Open Systems, France); Sébastien Pinneterre and Michele Paolino (Virtual Open Systems SAS, France)
The emergence of network function virtualization (NFV) has turned dedicated hardware devices (routers, firewalls, wide area network accelerators) to virtual machines (VMs). However, the costs to run these as virtualized network functions (VNF) in general purpose hardware is slower performance and higher energy costs. Field programmable gate arrays (FPGAs) are a promising solution to offset these performance and energy costs thanks to their reconfigurable fabric and acceleration capabilities. To offer quick access, high-throughput and sustain high-speed data streams many FPGA boards are bundled with the PCIe protocol. This protocol uses the Single Root I/O Virtualization (SR-IOV) standard to provide VMs direct access to the FPGA card and achieve close to native performance through hardware implemented virtual functions (VF). Thanks to SR-IOV VFs, VNFs can benefit of a simplified hardware exposure to the software. Despite that, SR-IOV has a strong impact on the resources usage flexibility and efficiency. As for flexibility, the number of VFs available is limited and this restricts the number of VNFs that can be accelerated. In fact, the VNFs that remain without a VF cannot take advantage of FPGA acceleration capabilities, even if there are FPGA resources available. In addition, if the VNFs that are attached to VFs underutilize the FPGA resource then the remaining VNFs that might be in greater need of acceleration are prevented from accessing the FPGA. In this paper, we present FPGA Virtualization Manager (vF- PGAmanager), an SR-IOV based framework that allows multiple VNFs to access different FPGA accelerators in a PCIe card through a single VF. vFPGAmanager is connected to these VFs through a multi-channel PCIe DMA and is responsible for routing packets to the corresponding accelerators. By using a multi- channel PCIe DMA vFPGAmanager is responsible for routing packets between VNFs and FPGA accelerators. Our key ideas are to containerize the VNFs inside a VM and provide them access to different accelerators in the FPGA fabric through the DMA and load balance the VF resource optimally. The results show a penalty of 30% in the worst case in throughput performance where two VNFs request acceleration at the same time from the FPGA using the same VF. For the best case scenario, when the VNFs don’t concurrently request FPGA acceleration, we have close to native performance.