Softwarization and Edge Computing Platform

Dec. 8th, 2020

Dai Kashiwa Evangelist, Director of NTT Communications ONF board member

• NTTʼs IOWN (Innovative Optical and Wireless Network) Concept

• Softwarization Challenges

• Edge Computing Platform

Estimating the amount of information Estimation of IT equipment power distributed through the Internet consumption [Billions of kWh] Double 5,500 200 121 Tbps 150 100 190 times 12 times 5,500 5,000 50 4,500 4,000 2,400 0 3,500 3,000 2,500 2,000 5 times 1,500 470 1,000 rate of trafficrate of growth 637 Gbps 500 0 2006 2025 2050 2006 2010 2015 2020 2025 Source: "Green IT Initiative" (2007.12) by the Ministry of Economy, Trade and Industry Source: "Green IT Initiative" (2007.12) by the Ministry of Economy, Trade and Industry Data growth Stagnation in technological evolution 175 ZB 180

160

140 120 90 times Operating frequency barrier 100

Zeta Bytes Zeta 60 40 202ZB Power consumption barrier 0 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

Source: https://www.karlrupp.net/2018/02/42-years-of-microprocessor-trend-data/ Source: IDC "November 2018 The Digitization of the World From Edge to Core"

"Low power consumption", "Large capacity, High quality" and “Low latency" networks by utilizing photonics technologies (Electronics to Photonics) Photonics-electronics convergence devices LASOLV Optical Processor •Quality guaranteed services by COSA assigning wavelengths to each service •Ultra-precise time information Optical transmission: •Fast computation of combinatorial 400 G optimization problems

connection All-Photonics around the chip Network

inter-chip optical transmission “Processing of photonics- Photonic layer electronics Low power Large capacity, Low latency convergence" consumption High quality through optical and electronic circuits Electrical efficiency Transmission End-to-end latency tight coupling capacity in-chip Electrical layer 100 times 125 times 1/200 intercore in-chip optical optical transmission signal processing Copyright ©NTT Communications Corporation. All Rights Reserved. IOWN (Innovative Optical and Wireless Network)

• Accurate prediction by simulation that takes into Digital Twin Computing account human thoughts and emotions (ex. Future (For new world of services and applications) city planning etc.) • Reproduction of realistic sensation beyond five C ( ognitive senses such as skin sensation (High realistic live, API For o CTL Edge Cloud expanded eSports, etc.) ptimally harmonize all ICT resources)

API overlay solution • Work allocation, decision making, future forecasts,

multiorchestrator etc., using your own self in cyber space

CTL

API network service F Operator A Operator C Agent Digital to Smart City Smart Sports Smart Workstyle Smart oundation Healthcare Operator B CTL Device Natural API Open Hardware + Software Electronics to

API All-Photonics Network Photonics (For significant improvement in the potential Smart World via IOWN of information processing infrastructure) ® Smart Smart Smart Smart Infra. Mobility Education Agri.

public network Private

Optical processor (photonics-electronics convergence technologies) https://iowngf.org terminal

Application

PaaS/SaaS

Infrastructure(IaaS)

Softwarized infrastructures are integrating into enterprises customers IT systems seamlessly...

Orchestrator Operation automation Controller

Predictive detection, automatic recovery

Value-added functions Easy upgrade Disaggregation Virtualizaion

Network OS Network OS Flexible configuration change

Open hardware Flexible scale up/down

Softwarization accelerates seamless integration with cloud-native ecosystems

CONVERGED ARCHITECTURES FOR NETWORK DISAGGREGATION & INTEGRATION NTT & Telefonica

PURPOSE • Define operator use cases in open converged packet and optical networks. • Prove that use cases can be met with architectures based on open technologies • Leverage the opportunity provided by TIP to involve different players to accelerate technical developments and help operators in real-world scenarios.

The target areas expand from the edge of the network up to the VNF or Datacenter platform going through the backbone network

Network EDGE VNF/DC Platform ODTN

VNF VNF Internet

Backbone Network

• Achieved in Madrid in Oc. and published white paper by 5 operators and ONF in Jan.

Remote Testing PoC Location: Telefonica Lab (Madrid) VPN NTT labs.(Tokyo)

Multi-Layer Controller (Metro-Haul project)

TAPI（ONF) Rest Optical Controller Packet Controller MSF (ONF: ONOS) (NTT:MSF) Netconf/ Netconf/Openconfig CLI Remarks CLI Openconfig TAPI（ONF) ・Carrierʼs S/W or OSS Controller OS NOS NOS OS NOS （NTT:Beluganos) Closed I/F Closed I/F TAI（TIP) Closed I/F TAI（TIP) OpenNSL Whitebox White-box White- ・Vendor Product Cassini (OCP) box White-box Open Line Transponder S/W （OCP) (TIP: Cassini) White- White- H/W Trans System box box Server ponder Transponder Server Multi Vendor Single Vendor Multi Vendor

[White paper on OOPT site] https://cdn.brandfolder.io/D8DI15S7/as/q43vmp-30aaqg-3sbb3u/CANDI_-1st_experimental_demonstration-_Telecom_Infra_Project.pdf

RU First Step 5G(NR)

Open Interface between RAN equipment CU/DU (i.e. decouple CU/DU from RU and 4G BBU) Open-Fronthaul

NTT DOCOMO has achieved multi-vendor interoperability Open-X2 of 4G and 5G base station equipment compatible with the O-RAN Alliance specifications. BBU 久野さん (e)LTE

Next Step Intelligence vCU/vDU introduction where it fits RIC based on requirements (with RIC) SW vCU/vDU (i.e. decouple HW and SW) HW(GPP) 5G(NR) RU Nationwide deployment of vRAN is not necessarily Virtualization efficient in terms of TCO today. NTT DOCOMO plans to (HW/SW decoupling) CU/DU deploy vRAN on the basis of requirements.

• Distributed processing platform to run value-added functions with optimal routes and environments dynamically, in cooperation with high-speed and low-latency underlay networks. • Build service meshes of value-added functions running on distributed multi-level edge environments • Support scale-in and scale-out of value-added functions and transfer them to other environments aligned with the running policies.

1~2ms 5~10ms 20~30ms 50ms~

10000+ locations 100+ locations ~20 locations ~10 locations

IoT Wireless UE GW Scale-out Home Core Wireline Device GW Network transfer

VPN Sensor uCPE Edge Platform Edge Platform

On-premises edge Access edge Core edge Cloud edge

• Centrally manage distributed edge platforms and remotely distribute value-added functions to specific edge platforms • Based on “Run Edge-Computing Everywhere” idea, from the Cloud edge to the On-premise edge, you can easily build an edge platform by using a global controller

deployment / Monitoring Centralize management (monitoring) of resource usage status by bundling bases

Global controller group１：factory

• Automatic location selection based on group２：Sales office properties of value-added functions • Remote version management of value- group３：50 people+ added fuctions sales office

Factory１ Factory２ Sales office３ Sales office４ Factory５

Off-loading mMTC slice massive Machine Type Communication Cloud Return by an edge Low-delay path selection uLLC slice Ultra-Reliable and Low Latency Communications Data Center

Wide-band path eMB slice selection enhanced Mobile Broadband CDN

• Provision of value- added functions such as network security, dynamic traffic control, tunneling, etc.. at the core edges • Open platform for our partners to enhance open collaboration

Network security, Dynamic traffic control, Tunneling, Dynamic deployment as... Load balancing,... network services or system integration Internet

AWS Azure GCP … Enterprise Cloud

Flexible InterConnect

Access Networks On-demand service chaining

Enterprise On-premises Core Edge

Distributing 5G functions according to characteristics • vRAN on on-premise edge -> UPF on Access edge -> 5GC on Cloud edge • A low delay is achieved by arranging the U plane processing unit at the edge Slicing WAN using SD-WAN Local breakout of traffic at the access edges and chaining to MEC

5G functions distrubting RU+DU + CU-U vRAN 5GC 5G AUS UD F M CU-C: RADIS Slice for APP1 Slice for APP1 C SD-WAN SD-WAN UPF

SliceInternet for APP2 Slice for APP2 Cloud Cloud Fabric Slice for APP3 MEC Slice for APP3 CU-C: NAT CU-C: L3

MPLS AM SMF PC Slicing WAN using F F SD-WAN Chaining to MEC On-premise Edge Acccess Edge Core Edge

Encryption/Decryption Load Balancing Core networks Core networks Message Routing Caching... Access edge Access edge

？ ×

Mobile access networks Mobile access networks

♪ ♪ ♪ ♪

ü Low latency communication ü Effective infrastructure resources utilization ü Dynamic and optimized route communications aligned with cars drive

IT before DX-era IT after DX-era 〜IT supports human works〜〜IT/human works together〜

Priority Cost, Performance Speed, Scale

Operations Fixed, Long-term, Stable Flexible, Agile, Continuous Delivery

Functions Process, Record Predict, Optimize

IP/ Know-how Patent, License, IP, Security Control Open Collaboration / Integration

Software speeds up innovation drastically and leads DX, integrating hardware evolution