Roberto Siagri presented at Eurotech's 45th Annual Meeting on accelerating technological change. He discussed how Moore's Law and human ingenuity have led to exponential increases in computing power over decades. Eurotech's strategy is to provide platforms that reduce customers' total cost of ownership through scalable software over scalable hardware. Siagri argued that emerging technologies like pervasive computing and the Internet of Things will continue advancing and becoming indistinguishable from everyday life through innovation.
3. 3
“The most profound
technologies are those that
disappear.
They weave themselves into the
fabric of everyday life until they
are indistinguishable from it “
The Computer for the 21st Century
"The Computer for the 21st Century", Scientific American,
Vol. 265 No.9, pp. 66-75, 1991
Mark Weiser
6. The Essence of Eurotech's strategy
Finding the right PLATFORM that reduce Customer’s TCO and TTM
It's a matter of SW vs. HW
Cost
Time
HW COST
SW COST
-
$
6
For Embedded
the PC
is the platform
choosen by
Eurotech
8. 20 years of progress
’92: A Personal Computer
on the palm
of a hand
’12: A Personal Computer
on half the palm
of a hand
16Mhz-1MB
1,6Ghz – 2GB RAM
8
X 10.000
10. Malthus was wrong. He forgot a factor:
our continual ability to do more and more with less and less because of
Technological Innovation
R. Buckminster Fuller
1895 - 1983
“The principle of doing ever more
with ever less
Space, Time, Matter
and Energy
per each given level of functional
performance”
Innovation moves from material to abstract
14. 14
Computer Food Chain : part 2
the new Computer Ecosystem
Pervasive
Computer
Computer Network/Cluster
Mini
ComputerMainframe
Vector
Computers
Parallel
Computers
Dave Patterson, UC/Berkeley
RIP RIP RIP RIP
15. Componentization
“
The rate of evolution of
any system is dependent
upon the organization of
its subsystems
”
Theory of Hierarchy and Componentization
Herbert A. Simon 1916-2001
17. APE project
1999: 0,1 TeraFlops per rack
QCD needs
Supercomputers
APEmille
2005: 1 TeraFlops per rack
APEnext
18. SISSA
MCA 2013
GPUs vs CPUs
comparing GPUs with
CPUs over the last
decade in terms
FLOPs, we see that
GPUs appear to be far
ahead of the CPUs
18
19. 19
Aurora Tigon: 140 TeraFlops per Rack
CPUs and GPUs power Aurora Tigon HPC cluster
to 3.2 GFLOPS/W
Sets World Record for Energy Efficiency
At the Top of the GREEN500
11000 CO2 tons saved
1500 cars that do not circulate for 1 year
11500 saved trees
15 Km2 of rain forest left untouched
20. SISSA
MCA 2013 20
Meta-Trends in Technological Acceleration
IDEAS are the new “ultimate” raw material
Moore's Law Miniaturization ‘65
Transistors increase 2 times over 18 months
Metcalfe's Law Interconnection ‘93
Value of a network increases with the square of the
number of connections
Gilder's Law Quantization ’00
Bandwidth increases 1,5 times over 12 months
21. 21
• Ubiquitous high bandwidth connection to the Internet at all times
• Massive computation available on demand through the CLOUD
• Tiny Computers embedded in
– the environment,
– our clothing,
– our body
Augmented real reality
Computers are becoming pervasive and ubiquitous
THE PLANETARY COMPUTATIONAL EXOSKELETON
22. 22
Cosmic Embryogenesis (in Three Easy Steps)
Geosphere/Geogenesis
(Chemical Substrate)
Biosphere/Biogenesis
(Biologic-Genetic Substrate)
Noosphere/Noogenesis
(Memetic-Technologic Substrate)
Le Phénomène Humain, 1955
The noosphere is a "planetary thinking network"
an interlinked system of consciousness and information,
a global net of self-awareness, instantaneous feedback
and planetary communication
Pierre Teilhard de Chardin
(1881-1955)
24. 24
What IoT needs
An universal infrastructure where data can exist
anywhere and be available to any device with
some form of consistency guarantee
Wearable
Notebook
Appliance
Sensor
Sensor
Server
Cluster
Network
Camera
25. IoT : i.e. from Embedded to Pervasive
computingCost
Time
HW COST
SW COST
-
$For the IoT
the Cloud
is the platform
choosen by
Eurotech
25
Finding the right PLATFORM that reduce Customer’s TCO and TTM
Again, it's a matter of scalable SW vs. scalable HW
26. “A Period of Combinatorial Innovation”
In the 1800s, it was interchangeable parts.
In the 1920s, it was electronics.
In the 1970s, it was integrated circuits.
In the 1990s, it was the VLSI functions.
In the 2010s, it was the SW Components.
Hal Varian
Google Chief Economist
27. The Internet of Things
Modern M2M Systems
Technology advances and standardization lead to a commoditization of
distributed systems
Where today’s solutions differ:
• Scaling up to larger applications & implementations
• Scaling down to smaller applications & implementation
• Interconnecting solutions on a platform level
• Much lower upfront investment
• Much faster implementation / deployment
• Much better TCO
28. APIs, Dashboards, Console, etc.
The Internet of Things
Eurotech Offer
Consumers of Data
M2M
Infrastructure
Solution
Technology
Building Block
M2M
Integration
Platform
Multi-
Service
Gateways
Producers of Data
Standard Interfaces
29. 29
Calm Technology
Mark Weiser
Ubiquitous computing just might
help to free our minds from
unnecessary work, and..
.. Connect us to the
fundamental challenge that
humans have always had:
to understand the patterns
in the universe
and
ourselves within them.
30. SISSA
MCA 2013
Addressing the increase in Software Complexity
Performance/Productivity
Complexity and Size
Assembly
Limited reuse
of written code
PC –Programming
Easy
A lot of available programs
Backward compatible
Cloud based apps
Multitenant
Mesh-up
Scalable
30
32. 32
Aurora & Aurora Tigon
Eurotech PETAscale HPC
Eurotech is part of
DEEP
the EU EXAscale
Research Project
33. 33
Growth
Time
today today + 3 years
One order of magnitude
change
“.. is really about human activity, it’s about vision, it’s about what you’re allowed
to believe… People are limited by their beliefs, they limit themselves by what
they allow themselves to believe is possible."
Carver Mead
EXPONENTIAL GROWTH
34. 34
1949: …a panel of expert
…. predicted that some day,
a computer as powerful as ENIAC would contain:
only 1,500 vacuum tubes; weigh 3,000 pounds;
and require 10 kilowatts of power to operate.
Such a machine would be about the size and
weight of an automobile with power consumption
to match.
Popular Mechanics
March 1949
35. 35
…we won't experience 100 years
of progress in the 21st century,
it will be more like
20,000 years of progress
(at today's rate)
Law of accelerating returns
Ray Kurzweil
The singularity is near
36. Sustainable Growth vs. Saturation
S curves get old
Time
Performance
Evolution phase
Take-off
Research,
Invention phase
Return reduction New technology
Existing
technology
New Evolution phase
36Each new TECHNOLOGICAL substrate is more efficient than the previous one
38. …but the computational
power per die is growing
Although single CPUs
have been limited, due
to the rise of multi-core
machines, the
computational power per
die has still been
increasing
38
40. SISSA
MCA 2013
1900 1920 1940 1960 1980 2000 2020 2040 2060 2080 2100
Calculationspersecond
The Age of Spiritual Machines (1999), by Ray Kurzweil
One insect brain
One mouse brain
1040
1035
1030
1025
1020
1015
1010
105
10
10-5
$1,000 of computing buys…
One human brain
All human brains
The Exponential Growth of Computing, 1900-2100
42. 42
Ultimate (zetta-wattaflop) Computing Machine
according to Seth Lloyd
a 1-kg computer compressed to the
black-hole limit can perform :
1051 ops/sec
on its: 1016 bits, updated at a rate of
about : 1035 update/sec
Seth Lloyd - “Ultimate physical limits to computation” Nature, 2000
Quantum Physics Theory
Super computers nowadays achieve
1015 ops/sec
with a pace of Moore’s law: x2 every 1.5 years
technology needs~ 250 years to reach these limits
43. The Benchmark…
~ 1016 ops/sec
25 Watt
1350 cm3
1500 gr
Surface
I7 CPU: 160 mm2
Brain : ~ 160 000 mm2
Assuming an I7 of the same surface
Operations per Second
I7 CPU: 10^11 Ops x 1000 = 10^14
Brain : ~ 10^16 Ops
Power
I7 CPU: 77 Watts x 1000 = 77 K Watts
Brain : 25 Watts
44. Computing Platforms in 2030
• Personal Computer [$1000]
– 10^10 Flops/sec in 2010 ⇒ 10^16 – 10^17 Flops/sec
44
• Supercomputer [$100,000,000]
– 10^15 Flops/sec in 2010 ⇒10^21 – 10^22 Flops/sec
• Number of Computers [global population ~10^10]
– SCs ⇒ 10^-8 –10^-6 per person ⇒ 10^2 – 10^4 systems
– PCs ⇒ .1x – 10x per person ⇒10^9 – 10^11 systems
– Embedded ⇒ 10x – 10^5x per person ⇒ 10^11 – 10^15 systems
– Nanocomputers ⇒ 0x – 10^10 per person ⇒ 0 – 10^20 systems
Available Flops Planet-wide
10^24 – 10^30 Flops/sec
[assuming classical models of computation]
45. Many Paradigm Changes
• Vertical Integration
– High core-cache bandwidth (new scaling roadmap for 15 years)
• Heat Removal
– Liquid cooling and heat re-use
• New Materials
– Carbon Nanostructured
• Supply and Cooling area vs. Transistors area
– Electrochemical power supply
• Production process
– 3D printing
45
46. Annual meeting American Physical Society 1959
Still … Plenty of Room at the Bottom
… But there is plenty of room to
make them [computers]
smaller.
There is nothing that I can see
in the physical law that says
the computer elements cannot
be made enormously smaller
than they are now.….
R. FEYNMAN
46
47. The only stable
component in nature
is the change!
The tentative fourth law of thermodynamics.
S.E. Joergensen
Thank you for your attention