5G Programmable In Infrastructure Converging dis isagg ggregated - PowerPoint PPT Presentation

5G-PICTURE 5G Programmable In Infrastructure Converging dis isagg ggregated neTwork and compUte Resources Anna Tzanakaki (University of Bristol, UK, NKUA, Greece)

Consortium • • Mellanox IHP GmbH (Coordinator) • Huawei Technologies • University of Bristol Dusseldorf GmbH • ADVA Optical Networking • Technische Universität • Airrays GmbH Dresden • Blu Wireless • Technology Transpacket • CNIT • Paderborn University • COSMOTE • COMSA INSTALACIONES Y SISTEMAS • EURECOM INDUSTRIALES SL • Fundació Privada i2CAT, Internet I • Innovació Digital a Catalunya Ferrocarrils de la Generalitat de • Telecom Italia S.p.A Catalunya • Zeetta Networks • • University of Thessaly 30 months duration • Start: 01/06/2017 • Universities (4x), Research Institutes (4x), SMEs (3x), • 7.99 million euro EU funding Operators (3x), Industry partners (5x) 2

5G-PICTURE Main Objectives • 5G-PICTURE will design and develop an integrated, scalable and open 5G infrastructure with the aim to support a variety of operational and end-user services for both ICT and “ vertical" industries . • This infrastructure will rely on a converged fronthaul and backhaul solution, integrating advanced wireless access and novel optical network domains. • 5G-PICTURE will rely on network softwarisation to enable an open reference platform instantiating a variety of network functions over a unified operating platform • slicing and service chaining to facilitate optimised multi-tenant operational models

Concept • 5G-PICTURE proposes to integrate network and compute/storage resources in a common infrastructure. • To address the limitations of current solutions 5G-PICTURE will adopt the novel concept of Disaggregated-Radio Access Networks (DA-RANs) • allows any service to flexibly mix-and-match and use compute, storage and network resources through HW programmability • relies on network softwarisation to enable an open reference platform instantiating a variety of network functions • adopts slicing and service chaining to facilitate optimised multi-tenancy operation • Hierarchical compute & storage structure supported by a network hierarchy • Integrated programmable wireless technologies at the edge and a hybrid passive/active optical transport network

Technical Approach • Τo address the limitations of D-RAN and C- e) 5G-PICTURE Monitoring and Profiling RAN will develop flexible functional splits Operations and Support OS System NFV Slicing (OSS) SDN Controller Orhestrator Manager • Adoption of the notion of DA-RAN relying on resource disaggregation d) PNF PNF VNF VNF • mixing-and-matching of resources (Synchronization) (Signal Processing/ VNF PNF massive MIMO) (vEPC) controller (beam tracking) VNF VNF VNF PNF (vMME, vCDN controller (FFT, iFFT) • Development of novel technology solutions vBBU) and control & management platforms VIM VIM VIM VIM c) b) • enhanced network and compute HW and Fast Fast Switch Switch module λ 1 module λ 1 BVT/ BVT/ BV- MUX SW modularity and flexibility BV- MUX R MPI R WSS MPI WSS Fast Fast Switch Switch module λ 2 module λ 2 OPP OPP MPI OPP OPP MPI • Creation and deployment of programmable BVT/ MPI MPI BVT/ BV- MUX BV- MUX R R WSS network functions and intelligent WSS Tx Fast Tx Fast Switch Switch module λΝ module λΝ orchestration schemes VIM VIM VIM VIM VIM VIM VIM • service chaining • slicing & multi-tenancy RRH mmWave PON RAM Storage Processing (specific) Processing (general) a)

Overall Infrastructure GPP: General Purpose Processor SPP: Specific Purpose Processor a) Elastic Optical Network Data Centre RU: Remote Unit Core BVT: Bandwidth variable transponder SPP SPP SPP RU 1 Node BV-WSS: Bandwidth variable wavelength ToR Edge selection switch Processing at the ServerRU Node SPP SPP SPP Processing at the RU Edge Optical Link RU 2 Node Fronthaul Flow Edge Core GPP GPP GPP  low network bandwidth b) Node Node ToR RU 3 Core  Limited BBU sharing MAC GPP GPP GPP Node (5) Decoding (4) Edge Node Receive c) ETH Tx λ 1 processing SFP RU Multi-protocol Interface I/Q Tx λ 2 (3) Fast Switch λ 1 BVT Fast ETH Tx Resource Switch Switch SYNC SFP RU module λ 1  high network bandwidth demapping λ 2 module λ 1 BV- I/Q Tx Multi-protocol Interface BVT λΝ  Increased BBU sharing BV- MUX WSS MUX WSS (2) ETH Tx Fast Fast SFP RU Cycle Prefix λΝ Switch Switch I/Q Tx Switch & FFT module λ 2 SYNC module λ 2 ETH Tx d ToR ToR SFP RU (1) ToR λ 1 RF to I/Q Tx SPP CPU Baseband ETH Tx λ 2 BV- MUX (2) SFP RU BVT BV- MUX SPP CPU Switch Switch WSS WSS I/Q Tx Tx Fast Fast SPP CPU Switch Switch (1) ETH Tx λΝ RU 2 module λΝ SPP SPP SPP SPP SPP module λΝ SFP RU SPP (3) (4) (5) I/Q Tx • Key Challenge: • Converged Fronthaul and Backhaul Services • Converged Network and Compute Services • Disaggregation of Resources 5

Physical Network Infrastructure RRH RRH a) Elastic Optical Network Data Centre Core RRH SPP SPP SPP RU 1 Node ToR Edge 5G-Xhaul Wireless Node SPP SPP SPP Edge RRH RRH 60GHz /Sub-6 links for FH/BH RU 2 Node Backhaul/Fronthaul FH Edge Core GPP GPP GPP Node Node ToR vBBU 1 vBBU 2 eNB RU 3 Core BH GPP GPP GPP Node Fronthaul eNB RRH λ   WDM-PON λ vBBU 1 λ λ λ λ λΝ RRH vBBU 2 λΝ λ λ TSON Small Macro vBBU λ RRH λ Cells Cell VM   λΝ λΝ λΝ Data Centers HeNB GW WLAN GW Backhaul HeNB vBBU WiFi HeNB S GW/GSN vBBU EPC Femto Cells vBBU HeNB Wireless Access Internet VM PDN-GW • Integrated optical and wireless network infrastructure for transport and access capitalising the 5G-Xhaul physical infrastructure • Wireless domain: • Dense layer of small cells complemented by macro cells to ensure ubiquitous coverage • Small cells can be backhauled to the macro-cell site either wirelessly using a combination of mm-Wave and sub-6 wireless technologies or using a hybrid optical network platform • Optical Domain • Adoption of a dynamic and flexible/elastic frame based optical network solution combined with enhanced capacity WDM PONs • BB Processing: RUs are connected to remote BB processing pools through high bandwidth transport links 6

Convergence of Networking and Computation VNF Mega DCs Internet Fronthaul adopting flexible • functional splits Core Resources Transport Network and • Controller VNF NFVI Mid-DCs Metro Metro Compute Resources VIM VNF Processing parallelisation • Regional Wireless Macro-cells DCs Access Resources Full GPP with commodity • Controller VNF NFVI VIM hardware and hybrid VIM Micro- VIM DCs solutions Resources VNF Resources Controller VNF Controller NFVI Disaggregation of Network and Frontlhaul NFVI • Compute resources PNF PNF Mixing and matching of RAM • RAM GPP SPP GPP SPP PNF SSD PNF VNF SSD resources to efficiently HWA GPP support services Remote/regional DC Cycle Prefix & FFT RF to Baseband demapping processing Resource Decoding Receive MAC & 8

Processing Parallelization Core Edge Node Node Edge a) Node RU 2 b) c) d) ToR ToR ToR ToR Switch Switch Switch Switch Switch Switch Switch Switch SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP SPP (4) (5) (4) (5) (4) (5) (4) (5) (3) (3) (3) (3) SPP/GPP SPP/GPP DC RU SPP/GPP RU SPP/GPP RU Optical inter DC network Intra DC Intra DC processing Decoding Receive Decoding Cycle Prefix demapping Cycle Prefix demapping Cycle Prefix demapping processing processing Baseband Baseband Resource Baseband Resource Decoding Resource MAC Receive network MAC Receive network MAC & FFT & FFT RF to & FFT RF to RF to Transport Transport Transport requirements: requirements: requirements: Split 2 (3) (4) (5) Split 4 (3) (4) (5) (3) (4) (5) Split 3 (1) (2) (1) (2) (1) (2) • BBU sub-frame processing time < 1ms • To reduce processing delay, BB processing is handled in parallel over disaggregated compute resources • Model 1: parallel (or sequential) processing mode • each function is distributed across multiple processing units (1:N) • Model 2: Pipelining • each processing unit handles a specific function adopting 1:1 mapping 9

Validation - Demo Activities Large DCs Internet SPP RAM GPP SSD Core RAM SPP GPP SSD Mid-DCs Metro Metro Wireless Mini DCs RAM Macro-cells GPP SPP Access SSD Micro- DCs • Extensive demonstration activities to showcase ICT and vertical industry use cases • Converged fronthaul and backhaul services in a smart city test-bed (city of Bristol, UK) • Seamless service provisioning and mobility management in high speed railway environments - 5G railway experimental testbed (FGC, Barcelona, Spain) • Media services supporting large venues with increased density and static-to-low mobility - stadium test- bed supporting large venues (Bristol, UK) 10 19/07/2017

UK 5G Trials Programme SDNx Switch & Cloud 5G & Fibre @ 5G & Big Data University of @ King’s Bristol & Bristol College is Open London Fibre (Janet) 5G Consumer Use- 5G Industry Use-Cases Cases 11 5G @ 5GIC, University of Surrey