Technologies that will Change the World

parsimoniousknotNetworking and Communications

Feb 16, 2014 (3 years and 5 months ago)

80 views


Click to add text

Technologies that will Change the
World

John Cohn Ph. D.

IBM Fellow

Systems and Technology Group


© 2011 International Business Machines Corporation

Our Industry has been driven by
Continual Improvement

and
Disruptive Innovation

!

Source: Kurzweil 1999


Moravec 1998

1900

1920

1940

1960

1E
-
5

1E
-
3

1E+0

1E+3

1E+6

1E+9

1E+12

$1000 Buys: Computations per second

Integrated

Circuit

Discrete

Transistor

Vacuum

Tube

Electro
-

Mechanical

Mechanical

2020

2000

1980

© 2011 International Business Machines Corporation


Exascale

(Datacenter
-
in
-
a
-
box)


Massive parallelism


Flexible system
optimization

Nano
Devices

1B Transistors

Power7 chip

1T Devices

Nano Systems

(Systems
-
on
-
a
-
chip)

Workload
Optimized
Systems

Big Data


Compute+


Natural


Language+


Analytics

Cognitive Computing


“Synapse” devices


Photonics


DNA Transistor


BIG
/
Fast


Data + analytics
(zettabytes +


milli / microseconds

Deep Q&A
Computers

4

Technologies that Will Change the World



and IBM Will Lead !

© 2011 International Business Machines Corporation


Exascale

(Datacenter
-
in
-
a
-
box)


Massive parallelism


Flexible system
optimization

Nano
Devices

1B Transistors

Power7 chip

1T Devices

Nano Systems

(Systems
-
on
-
a
-
chip)

Workload
Optimized
Systems

Big Data


Compute+


Natural


Language+


Analytics

Cognitive Computing


“Synapse” devices


Photonics


DNA Transistor


BIG
/
Fast


Data + analytics
(zettabytes +


milli / microseconds

Deep Q&A
Computers

From Nano Devices to Nano Systems

© 2011 International Business Machines Corporation

22/20 nm

15/11 nm

8 nm & Beyond

Conventional
Planar Device

FINFET

ETSOI

Si Nano
-
Wire

C Electronics

Fully Depleted
Devices

Si NW

HfO
2

Deposited Si

Device Structure Research Pipeline

Innovation and Disruptive Technology at Each Node

© 2011 International Business Machines Corporation

Light out

for off
-
chip traffic

Modulator

Photo
-
detector

Analog
CMOS

Digital
CMOS

Inter
-
layer

3D vias

Multiplexor

Vision: >1 Tbps on a 3D Chip

© 2011 International Business Machines Corporation

The Future of Medicine

“Lab
-
on
-
a
-
Chip”

“DNA Transistor”


Requires pinhead size sample; takes only minutes


Tests for hundreds of diseases, viruses, bacteria and
biohazards


Goal: a personal genome for $1,000


Could make possible customized medicine

“Targeted Medicine at the Nanoscale”



Detects and destroys antibiotic
-
resistant bacteria



Fights infectious diseases like MRSA

© 2011 International Business Machines Corporation


Exascale

(Datacenter
-
in
-
a
-
box)


Massive parallelism


Flexible system
optimization

Nano
Devices

1B Transistors

Power7 chip

1T Devices

Nano Systems

(Systems
-
on
-
a
-
chip)

Workload
Optimized
Systems

Big Data


Compute+


Natural


Language+


Analytics

Cognitive Computing


“Synapse” devices


Photonics


DNA Transistor


BIG
/
Fast


Data + analytics
(zettabytes +


milli / microseconds

Deep Q&A
Computers

From Petascale to Exascale

© 2011 International Business Machines Corporation

BG/L

0.3PF

BG/P

1PF

Roadrunner

1PF

~300PF

~1000PF

z9

z10

zEnterprise

pSeries

P5

pSeries

P6

pSeries

P7

BG/Q

~20PF

Relative Performance (log)

Time

From Disruptive HPC to Commercial Leadership

© 2011 International Business Machines Corporation

1 PetaFlop = 1/3 rack

Software

Phase

Change

Memory

CPU

Silicon
Photonics

3D

1 PetaFlop

72 BG/P Racks

Overall Performance = 1000X


Performance / watt = 135X


Performance / $ = 1000X


Footprint = <2%



Referenced to one
-
petaflop system

The Next Ten Years

2009

Exascale Computing

A billion calculations in a billionth of a second

2019

© 2011 International Business Machines Corporation


Exascale

(Datacenter
-
in
-
a
-
box)


Massive parallelism


Flexible system
optimization

Nano
Devices

1B Transistors

Power7 chip

1T Devices

Nano Systems

(Systems
-
on
-
a
-
chip)

Workload
Optimized
Systems

Big Data


Compute+


Natural


Language+


Analytics

Cognitive Computing


“Synapse” devices


Photonics


DNA Transistor


BIG
/
Fast


Data + analytics
(zettabytes +


milli / microseconds

Deep Q&A
Computers

From Big Data to Big Analytics

© 2011 International Business Machines Corporation

Smarter Planet will Drive Creation of
Big
/
Fast

Data

Millions of end points (e.g., meters, appliances, capacitors)

Analytics requirements: 100K+ data elements,10ms latency

Multiple feedback time
-
scales: sub
-
second, 4 sec, 5 min, …

Intelligent Building

Demand Response

5 minute price signals

Electric Car

Charge Points

430 locations

Substation

Switch Gear

40 control points

Transmission Line

SynchroPhasor Sensors

30 measurements / sec

Automated Meters

Households

5.2 million

Feeder Line

Power Quality

Sub
-
second feedback

© 2011 International Business Machines Corporation

Telco Promotions

100,000 records/sec, 6B/day

10 ms/decision

270TB for Deep Analytics

DeepQA


100s GB for Deep Analytics

3 sec/decision

Smart Traffic

250K GPS probes/sec

630K segments/sec

2 ms/decision, 4K vehicles

Homeland Security

600,000 records/sec, 50B/day

1
-
2 ms/decision

320TB for Deep Analytics

Traditional Data Warehouse &
Business Intelligence

Data Scale

yr

mo

wk

day

hr

min

sec



ms


s

Exa

Peta

Tera

Giga

Mega

Kilo

Decision Frequency

Occasional

Frequent

Real
-
time

Up to
10,000
Times
larger

Up to 10,000 times
faster

Data in Motion

Data at Rest

New
Big
/
Fast

Data Brings New Opportunities,

Requires New Analytics

Big

Fast

Big Insights

System s

Feedback

© 2011 International Business Machines Corporation


Exascale

(Datacenter
-
in
-
a
-
box)


Massive parallelism


Flexible system
optimization

Nano
Devices

1B Transistors

Power7 chip

1T Devices

Nano Systems

(Systems
-
on
-
a
-
chip)

Workload
Optimized
Systems

Big Data


Compute+


Natural


Language+


Analytics

Cognitive Computing


“Synapse” devices


Photonics


DNA Transistor


BIG
/
Fast


Data + analytics
(zettabytes +


milli / microseconds

Deep Q&A
Computers

From Programming to Systems that Learn

© 2011 International Business Machines Corporation

Watson !


A Workload optimized system


Built on 90 x IBM Power 750 servers


2880 POWER7 cores


15 Terabytes of memory


Up to 80 Teraflops


500 GB per sec on
-
chip bandwidth

Hypothesis

Generation

Evidence

Scoring

Confidence
Ranking

Solution

Synthesis

Question
Analysis

Problem

Decomposition

© 2011 International Business Machines Corporation

Watson


a Roadmap

2007


2011

2011


2012

2012


2015

Research / Demo

Commercialization

Future Technologies

Won
Jeopardy!

Voice & Image Recognition

Query & Dialogue

© 2011 International Business Machines Corporation

Deep Q&A: Progress on Medical Queries

Watson

Y/E 2010

Starting Point

2007

% Answered

100%

0%

0%

100%

Precision

“Winners Cloud”

% Answered

100%

0%

0%

100%

Precision

DeepQA

Adapted for Medical

Questions

~3 person
-
month

effort

Eg. This is the most common
cause of unilateral decreased
vocal fremitus.

© 2011 International Business Machines Corporation

Neuroscience

Supercomputing

Nanoelectronics

Complexity

Time

Cognitive Computing

© 2011 International Business Machines Corporation

Computer Intelligence

Time

Tabulating
Era

Computing Era

Smart

Systems Era

We Are Entering a New Era

© 2011 International Business Machines Corporation