internet of programmable things for in-network data analytics
TRANSCRIPT
Internet of Programmable Thingsfor In-Network Data Analytics
All Rights Reserved by Akihiro Nakao, 2018 1
The University of TokyoAkihiro Nakao
All Rights Reserved by Akihiro Nakao, 2018
Marriage betweenSoftwarization and Data Analytics
2
• Flexible infrastructure for sensing/analyzing/learning traffic in data plane
Data Plane ProgrammabilityIn-Network Machine Learning
• Data analytics for complex actuation (automated agile operation) human may not achieve
All Rights Reserved by Akihiro Nakao, 2018
Mobile Network Prediction for 2021Visual Networking Index (VNI)
The number of mobile users
The number of mobile terminals connected
7x
In 2021, mobile traffic will amount to 48.3 EB per month1EB = 1018B
Global increase in mobile network traffic
Cisco Visual Networking Index
Global annual data center traffic 1ZB=1000EB = 1021B
Global increase in data center traffic
3
All Rights Reserved by Akihiro Nakao, 2018
Data Type / Field Project Summary
Location Information NTT Docomo“Mobile Spatial Statistics”
Provide population statistics from the anonymized location data of mobile phones
Automobile Probe Toyota“Telematics Service”
Provide traffic information and statistics generated from telematics data for improving traffic congestion and public safety
Automobile Probe Sony Assurance Inc.“Telematics Insurance”
Analyze customers’ telematics record and provide cash back for safety driving
Medical Information NTT Docomo Health-Care”Moveband3”Omron Health-Care“Wellness Link”
Provide services for improving health and life style by visualize and analyze activity data obtained from wearable smart wrist bands.
Financial Information HitachiFinancial API Service
Enable personal asset management across multiple financial accounts
http://www.soumu.go.jp/johotsusintokei/whitepaper/ja/h29/pdf/n2100000.pdf
Utilization of Mobile / IoT Data
4
All Rights Reserved by Akihiro Nakao, 2018
WP1�Coordination & Management (AALTO, UT/NESIC) WP6�Promotion and Standardization (All Partners)
Multi-RAT Front/Backhaul Core Network Data-center Network
Sensor Device Network
Network Resources
Computing/Storage resources Mobile Edge Computing
Terminal/ Device Data-center
WP2�Network Softwarization Architecture (All Partners)
WP5�Federated Testbeds and Experiments (UT, WU, NESIC, AALTO, Ericsson, Orange, FOKUS)
Data center Base station
WP4�E2E Slice Orchestration (UT, KDDI, HITACHI, AALTO, Ericsson, Orange)
Orchestration
Creation of specific slice for each service
Advanced ITS Slice
WP3�Network Slicing Mechanisms (UT, WU, NESIC, Ericsson, AALTO, FOKUS, Orange)
CDN/ICN Slice
IoT/M2M Slice
10-100Gbps broadband/new protocol oriented
1ms Low- latency
106 devices/km2 oriented
Operation of virtual network for different mobile services
Operation of virtual network for different mobile services
5G/IoT Service Centric Network Slicing Control and Operations over Multi-Domains and Multi-Technologies Future Infrastructure
10-100Gbps broadband/new protocol
1ms Low-latency
106 devices / km2 oriented
R&DStatementandGoals1. WP1: Coordination and Management
Project management through iterating PDCA cycles
2. WP2: Network Softwarization Architecture Development of unified E2E Mobile Infrastructure Architecture enabling multiple hundreds of slices instantiation
3. WP3: Network Slicing Mechanisms Development high-performance and highly programmable C/D-plane execution platform Development of ICN, advanced protocols, over slices
4. WP4�E2E Slice Orchestration Development of distributed schemes and platform for achieving scalable operations and management
5. WP5�Federated Testbeds and Experiments Technical integration and federation of technical assets of WP2-3, and PoC systems development and evaluation
6. WP6: Standardization Contribution and leadership in Global SDO, e.g. ITU, 3GPP
TechnicalGoalsStandardization and Verification of E2E Network Slicing Technologies through EU/Japan Collaborative R&D efforts (1) Network Softwarization and E2E
Network Slicing Architecture (2) Data Plane Programmability and its
Advanced Networking Protocols Instrumentation
(3) Scalable E2E Slice Operations and Managements (Orchestrations)
Blue-PrintofE2ENetworkSlicing
Anetworksliceforeveryservice!
EU-Japan Jointly Funded Project on 5th Generation Mobile Network (PIs: Akihiro Nakao@Utokyo and Tarik Taleb @Aalto University)
5G!Pagoda is funded by the European Commission’s H2020 program under grant agreement n° 723172.
EUJ-01-2016 - 5G – Next Generation Communication Networks
EU Total cost:EUR 2.2MJP Total cost:225 M JPY
Funding Size
Duration: 3 years(2016-2019)
5
All Rights Reserved by Akihiro Nakao, 2018
https://www.u-tokyo.ac.jp/adm/fsi/en/sdgs.html
UTokyo FSI promotes SDG-oriented projects in a wide range of fields throughout the University, and showcases them as actions taken by the University as a whole.In particular, in regards to collaboration with the industrial sector, the University utilizes the SDGs as a basic common vision for new business growth.
Sustainable Development Goals (SDGs) as Common Vision
As of 2018/4/5, 170 SDGs projects have been registeredhttps://www.u-tokyo.ac.jp/adm/fsi/ja/projects.html
6
All Rights Reserved by Akihiro Nakao, 2018
SoftwarizedProgrammable UE/Sensors
SoftwarizedProgrammableeNB/EPC/MEC
“Softwarization Everywhere”= E2E Programmability
SoftwarizedCloud Data Centers
DataAnalyticsPossibility
DataAnalyticsHeavily Conducted
DataAnalyticsPossibility
DataAnalyticsPossibility
SoftwarizedProgrammable IoT GateWay
7
All Rights Reserved by Akihiro Nakao, 2018
Sliceable Software Defined Data Planes
8
Control-PlaneElements
Network Applications
ProgrammableData-PlaneElements
Applications
Control Plane
Data Plane ProgrammableData-PlaneElementsProgrammableData-PlaneElementsProgrammableData-PlaneElements
AI/ML
AI/ML
• ML Offload• Autonomous OAM• Annotation• Sampling• Characteristics Extraction
• Deep ML• AI OAM
All Rights Reserved by Akihiro Nakao, 2018
In-Network Machine Learningfor Application Identification
9
All Rights Reserved by Akihiro Nakao, 2018
C C A A
CC A C
CDAFCA C
AC A
AC
D A
DAA
CDA FCA C C C C
A C
E C
CDA FCA C
CDAC
A
A C
CDAC
AA
S N P M
DC
IV R LWTU OGW
CE A A CAD CDA
CE A
C C
JHPCN Project JH170041-NWH @ NakaoLabThe Joint Usage/Research Center for Interdisciplinary Large-scale
Information Infrastructures
C
All Rights Reserved by Akihiro Nakao, 2018
Application-Specific Bandwidth Control(FLARE and P4)
Youtube (mediaserver): 3Mbps
android.browser: 500kbps
Chrome 2Mbps
All Rights Reserved by Akihiro Nakao, 2018
Association betweenLow-Level Operation Data
andHigh-Level Application Identification
13
All Rights Reserved by Akihiro Nakao, 2017
Softwarized Base Station For Data Analytics
RRCPDCPRLCMACPHY
eNB
L1
L2
L3 CPU
DSPFPGA
Offload complex processing To FPGA and DSP
RRCPDCPRLCMACPHY
CPU
VM
eNB EPC
VM
SP-GWMMEHSS
MEC
VM
Sharing H/W resources among eNB, EPC and MEC
service
SoftwarizationOf all digital
Signal processingn Challengesü Processing power for complex data handlingü Stable operation especially fluctuation in latencyü Sharing computational resources with network function virtualization
Conventional Approach Our Approach
14
DataAnalyticsPossibility
DataAnalyticsPossibility
DataAnalyticsPossibility
All Rights Reserved by Akihiro Nakao, 2017
OAI Field Experiment on Hongo CampusCollaboration with Fujitsu
�
�
�
����
Aki Nakao,, “R&D on End-to-End Network Slicing”, OAI Workshop Paris 201715
Experimental License @1.7GHz
All Rights Reserved by Akihiro Nakao, 2017
LTE Softwarized Base Station
Attach to LTE
Voice Call
Video Streaming Data DL
ü 5510 ü 2ü 2ü 5 45 + 1 4 5 + 1
17
All Rights Reserved by Akihiro Nakao, 2018
Preliminary Experimental Results
DL: 67.27MbpsUL: 13.88Mbps
20MHz5MHz
DL: 8.44MbpsUL: 7.99Mbps
10MHz
DL: 14.91MbpsUL: 12.44Mbps
• The experimental network is deployed on LTE-FDD Band3 (1.7GHz) to avoid interfering with other carriers (Docomo/au/softbank). • Throughput was measured with SpeedTest Application running on a Fujitsu’s Arrows M04 Android phone.
18
All Rights Reserved by Akihiro Nakao, 2018
Collecting Operation Data in Real Time
n 0 4 0 0 44 04 04
0 40 04 4 0 4 4 04
19
All Rights Reserved by Akihiro Nakao, 2018
GDPR
What constitutes personal data?Any information related to a natural person or ‘Data Subject’, that can be used to directly or indirectly identify the person. It can be anything from a name, a photo, an email address, bank details, posts on social networking websites, medical information, or a computer IP address.
22https://www.eugdpr.org/eugdpr.org.html
All Rights Reserved by Akihiro Nakao, 2018
50% of citizens share data by 2019: Gartner
Predicts 2017: Government CIOs Are Caught Between Adversity and Opportunity
Gartner predicts that by 2019, 50 percent of citizens in million-people cities will benefit from smart city programs by voluntarily sharing their personal data.
https://www.canadianunderwriter.ca/keyword/predicts-2017-government-cios-are-caught-between-adversity-and-opportunity/https://www.gartner.com/doc/3510217/predicts--government-cios-caught
23
All Rights Reserved by Akihiro Nakao, 2018
Challenges• Flexible (data plane) Infrastructure• Machine Learning Offload• Autonomous OAM• Annotation• Sampling• Characteristics Extraction• In-Network Machine Learning • Sensing / Inference without privacy violation• Operational Data• Traffic Data• Social Network Application Data• Viable Use Cases
24