The Edge: Evolution or Revolution? Computing SEC Keynote Oct...November 2016 The Edge: Evolution or Revolution? Pablo Rodriguez, CEO alpha.company October 2017 ACM SEC, Symposium on

November 2016

The Edge: Evolution or Revolution?

Pablo Rodriguez, CEO

alpha.companyOctober 2017ACM SEC, Symposium on Edge ComputingSan Jose, California

2Pablo RodriguezTelefónica Digital

Internet Design

Lacked:• Mobilty• Security & Privacy• Economics Sense that the Internet was free

• Content Distribution

2

P. Rodriguez, “Scalable Content Distribution in the Internet”, Ph.D. Thesis, EPFL 1997-2000

4

Overlays Evolution

Hype

Realism

Growth

CachingIP Multicast

CDNsAkamai

Enterprise CDNsEdge Insertion

Layer-7 SwitchesSatellite CDNs

Video

1999

2000

2001

2002

2003

2004

Disappointment

Computing

P2P

5

My Upbringing• EPFL/Eurecom• Inktomi, Adquired by Yahoo!• Netli, now part of Akamai• Tahoe Networks, now part of Nokia• Bell-Labs• Microsoft Research • Telefonica

Impact

research

prototype

launch

6

Three Edge Example projects

• 1. Wireless Proxies (Bell Labs)• 2. P2P Content Distribution (MSFT)• 3. Video CDN (Telefonica)

Large File

Two peers,left peer is twice faster

6 pieces1 2 3

654

1 2 3 4 5 6

Client

Content

IP

BSS

Core Network

GGSN W-PEP

Application Optimizations(e..g. compression)

TCP Optimizations

(e.g. connections sharing, ACK regulator)

MAC optimizations

(e.g. Qos, FEC, scheduling)

Session Optimizations(e.g. DNS Boosting)

(1)

(2)

(3)

Windows Updates Telefonica Video CDN

Wireless Proxies

Telefónica Alpha / Edge Computing / 7

1Back to the Future

8

The Problem• The Internet has been growing very fast, both in

the number of users and in the available content– Overloaded servers and network links.– Frustrated users.

The World Wide “Wait”

“The Web can kill the Internet”, B. Metcalfe 1995

9

Not ready for Flash-Crowd Events

10

P. Rodriguez, “Scalable Content Distribution in the Internet”, Ph.D. Thesis, EPFL 1996-2000

11

Centralized

Origin Server

Caches

Name: www.foo.com IP: 192.12.12.5

12

Hierarchical

Origin Server

Caches

Name: www.foo.com IP: 192.12.12.5

Hierarchical Caching Overlay

P. Rodriguez, E. Biersack, “Web Caching: Hierarchical vs Distributed Caching”, IEEE/ToN 1999

13

1 2 65

P2P File Distribution

Server

3 4

1 5 6 2 4

1 2 3 4 5 6

3

T. Karagiannis et al., “Planet Scale Software Updates”, Sigcomm2006. Windows Updates

P. Rodriguez, E. Biersack, “Parallel Access for Mirrors in the Internet”, Infocom 2000

14

Nano Data Centers

Home Edge Cloud functionality

17

THE CLOUD

November 2016

Univac video

23Telefónica Alpha / Edge Computing /

Connected devices have never collected so much data about their environment (vision, location, temperature...). That real-world data is massive(e.g. a self-driving car generates about 10 gigabytes per mile).Pushing it back to the cloud will become increasingly difficult. Existing infrastructures will not be able to handle its volume.

IoT objects become more and more sophisticatedTHE ADVENT OF EDGE COMPUTING

UGC IoT


1. An enormous amount of data is already being generated

10 Gb / mile 400 Gb / secLytro Cinema VRAutonomous Car

THE ADVENT OF EDGE COMPUTING


Two trends that will disrupt the cloud paradigmTHE ADVENT OF EDGE COMPUTING

Increasing data Real-time needs


Challenge #2: How many Edges?

1-20 20-80milliseconds milliseconds

1

Visual-tactile feedback

ms Human Features

ms

Network & Spatial

ms

Use CaseRequirements

1

1

Tactile Internet

5G (target)

1

Car at 140 km/h, 3.9 cm

Wearable Assistance

4

Tell that a sound is a human

voice

7

Vestibulo-ocular reflex

Visual reaction time

1

Speed of light per 200 km

5

Augmented Reality (gaming)

7

Autonomous Driving

5

Two people in a room, latency added by speed

of sound

8

Car at 90 km/h, 20 cm

TV max inter picture

Virtual Reality

20

WiFi (average)

20

An airbag to release

30

RTT UK N-S (IP) (564 miles)

30

Perception of a screen change after clicking a

button

50

4G/LTE (average) 70

RTT USA W-E (IP) (2514 miles)

80

Screen to brain propagation

100

Auditory reaction time

130

3G (average)150

Human blink

150

Crisp UI with a computer

80

Augmented Reality

(non gaming)

100

TeleRobotics

Telefónica Alpha / Edge Computing / July 2017

10

10

10

10

80-150milliseconds

>150milliseconds


Many edges, per use case and workload

November 2016

Evolution of Video traffic (Global)

Peak Internet traffic will grow at a compound annual growth rate of 35% from 2016 to 2021, compared to 26% for average Internet traffic.

CDNized content will grow to 70% globally

+20pp

Incremental Edge Computing Examples

31

Virtual CDNs at the EDGE

32

Propriety and ConfidentialProduct Innovation . Customer Centric Networks

Access Aggregation Core Peering

Access

AccessAggregation

Aggregation

Aggregation

X.000.000 X00 X0 2-4>10

CDNs Today…

33


Access

Aggregatio

nCore Peering

Access

Access

Aggregation

Aggregatio

n

Aggregation

X.000.000 X00 X0 2-4>10

CDNs tomorrow… Virtual CDNs at the Edge

34


The connected home of the past

InternetCPE

The connected home of today

http://www.ericsson.com/res/docs/2014/virtual-cpe-and-software-defined-networking.pdf

CPE

IPv4 NAT

SwitchAccess Point ModemDHCP

FWTR-069 UPnP

STB

SwitchAccess Point Módem

CPE FW

TR-069

NAT

UPnP

DHCP

IPv4/IPv6

STB

FROM…

… TO

Home environment

Home environment

Network environment

Network environment

Simplification removes all incompatibilities

New functions (e.g. IPv6) only needed in network environment

Operation and service deployment are greatly simplified

Innovation in services: The vCPE Principle

Credit: Antonio Elizondo, Diego Lopez, Telefonica GCTO


Challenge #3: Where is the “killer app”?THE ADVENT OF EDGE COMPUTING

Edge computing is still in its infancy and a framework to facilitate its adoption is not yet available.

Some use cases are only emerging (e.g. VR/AR will be a niche for some time and industrial IoT is a very narrow vertical).

How to create the market demand? Nobody will develop apps that require


The killer app will come from the sweet-spot between bandwidth & latency.

latency

bandwidth throughput1Gbps

1 ms

1000 ms

100 ms

10 ms

Virtual Reality

Tactile Internet

Autonomous DrivingAugmented

Reality

Most suitable-for-the-edge use cases are heavy on video and computer vision

Emerging Robotics



First, it’s the AI, stupid.ENABLING EDGE COMPUTING

AI-powered applications will catalyze the adoption of edge computing. It is all set to become the most preferred architecture for running data-driven, intelligent applications.

IDC estimates that today, only 1% of application across all industries have some type of cognitive technology. In two years that number will exceed 50%.

Edge computing prioritizes agility over power. Endpoints will never be as powerful as the cloud can be. On the other hand, they gain agility from the speed of the information loop that occurs in the edge, processing just the information that is needed.

The cloud will then become a place where learning happens.


AI Fields and Applications

Use input from sensors to deduce aspects of the world. Computer vision. Speech, facial and object recognition.

Recognise, interpret, process, and simulate human affects, emotions and social skills.

SENSE

INFER

ACT

Perception

Social Intelligence

Representation of "what exists”, qualification and commonsense.

Read and understand the languages that we speak. Machine translation.

Systems that identify/assess creativity or generate outputs that can be considered creative.

Knowledge Representation

Natural Language Processing

Creativity

Robots to be able to handle such tasks as object manipulation and navigation.Motion and Manipulation

Make logical deductions dealing with uncertain or incomplete information.Reasoning and

problem solving

Algorithms that improve automatically through experience.

Set goals and achieve them. Also in cooperation (swarm intelligence).Planning

LEARN

ENABLING EDGE COMPUTING


Secondly, specialised hardware will be neededENABLING EDGE COMPUTING

To enable edge computing, besides AI processors and algorithms, there’s the increasingly important task of creating engineering systems to maximise performance.

New and specialised chips and systems are needed to take AI to the next level. Boutique chips will be developed to deliver better performance and massively reduce training requirements and improve costs. The objective is to build computational platforms that deliver the performance and energy efficiency needed to build AI with a maximum level of accuracy.


3Data Control at theEdge

51

“Biggest risk for the Web: Losing control of Data”,

Sir Tim Berners-Lee 2017

Personal Data at the Edge is far more valuable than aggregated data. The more private/intimate and the closer to the context of the user, the more valuable it becomes.

$0.10 $0.18 $0.62 $6.50 $32.15 $54.50

Anonymous profile

with 1 identifier

Anonymous profilewith 34

identifiers

Value per "friend" per

Facebook profile

Location Data per

contact in DB

Demographics data

per contact

Buying Behavior &Preferences

Increase in value ofcontact as it becomes attached to an identity

x2 x6 x65 x320 x545

Source: Atos Group

Papadopoulos et al., “If you are not paying for it, you are the product”, IMC 2017

Revenue comparisons

World GDP = $70,000 BTelco Revenue (>50% Wireless) = $2,000BPersonal Data (Ads) Revenue = $500BVideo/TV Revenue = $182BTransit Revenue (Level 3) = $6 BAkamai 2011 Revenue = $1.2 B

59

Edge Functionality Evolution

Application Improvements(e.g. Compression, Caching, Edge insertion, Cyber Security)

Protocol Optimizations(e.g. Delay-jitter algorithm, ACK regulator, Pacing)

Analytics/Algorithms(e.g. Vision, AI/ML as a service, Big Data, Cognitive Decision, Situational awareness)

Privacy/Data Control(e.g. IID protection, Data Control, Anti Tracking, Transparency, Data Banks)

EDGE TREND: In-Network Privacy Control Functionality

PARENTAL FILTERMALWARE

VIRUSAD MANAGEMENT

PRIVACY DATA PROTECTBLOCK OF TRACKING

IDENTITY TRANSLATION-PROTECTION

Naylor et al, “McTLS: Enabling secure in-network Funcionatlity in TLS”, Sigcomm 2015

Personal Data: Give Data and Value Back

Individuals leave data traces (location, calls, web traces, shopping, etc)Such info is useful for Credit Ratings, Retail Industry, Govs, Online

Advertising, Predictions and Analytics (Sexual Orientation, Political Views)Individuals have multiple virtual Data Souls and PersonalitiesIndividuals with access and control of that data could use it to their benefitWe still live in a “data desert”: we don´t know much about ourselves or the

world (where do people like me go, where is it safe, where can I find a job)Give value back to our customers

- “For sale: Your Data: By you”, Hotnets 2011

Erramii et al, “For sale: Your Data: By You”, Hotnets 2011

Personal Data Banks

- “My Data Soul”, P. Rodriguez TEDx Talk- www.rodriguezrodriguez.com


4 Main Takeaways


It’s about moving computation and storage closer to where data is created and acted uponIt’s so far being transformationalThere will be many edges

The disruptive opportunities will come from use cases that require high bandwidth and low latency; mainly video and computer vision.

The Edge will enable AI in Real Time as a service

Specialised Hardware will be the catalyser

It will require really huge investments and cross industry strategic partnerships with sticky relationships. It will not happen by chance.

Privacy, Security and Data Value and control will drive a large percentage of Edge use cases

The edge in a nutshellMAIN TAKEAWAYS

alpha.company

November 2016

Slide Number 1Internet DesignSlide Number 3Overlays EvolutionMy UpbringingThree Edge Example projectsSlide Number 7The ProblemNot ready for Flash-Crowd EventsSlide Number 10CentralizedHierarchicalP2P File DistributionSlide Number 14Slide Number 15Slide Number 16THE CLOUDSlide Number 18Slide Number 19Slide Number 20Univac videoSlide Number 22Slide Number 23Slide Number 24Slide Number 25Slide Number 26Slide Number 27Slide Number 28Slide Number 29Slide Number 30Slide Number 31Slide Number 32Slide Number 33The connected home of the pastThe connected home of todayInnovation in services: The vCPE PrincipleSlide Number 37Slide Number 38Slide Number 39Slide Number 40Slide Number 41Slide Number 42Slide Number 43Slide Number 44Slide Number 45Slide Number 46Slide Number 47Slide Number 48Slide Number 49Slide Number 50Slide Number 51Slide Number 52Revenue comparisonsSlide Number 54Slide Number 55Slide Number 56Slide Number 57Slide Number 58Edge Functionality EvolutionSlide Number 60 EDGE TREND: In-Network Privacy Control FunctionalityPersonal Data: Give Data and Value BackSlide Number 63Slide Number 64Slide Number 65Slide Number 66Slide Number 67

Documents

The Edge: Evolution or Revolution? Computing SEC Keynote Oct...November 2016 The Edge: Evolution or Revolution? Pablo Rodriguez, CEO alpha.company October 2017 ACM SEC, Symposium on