This talk is based on the following material:
A. Liotta, G. Exarchakos
'A Peek at the Future Internet'
Springer (2011)
http://bit.ly/FI_liotta
A. Liotta
'Cognitive Interconnections'
ISBN 978-90-386-2518-8 (2011)
http://bit.ly/booklet-antonio
Future Internet: how diverse disciplines will help redesigning our networks
1. Future Internet:
How diverse disciplines will help
redesigning our networks
Prof. Antonio Liotta
Eindhoven University of Technology, NL
http://bit.ly/autonomic_networks
Twitter: a_liotta
2. What does it take to be a ‘generative’ Internet?
• Apps: ~0.5 million, 600 new
apps/day
• Multicast IPTV: $38m
expected revenues for 2013
• Internet video: 0.8 Zettabytes
(1021bytes) in 2014
• Connected ‘things’: 1 trillion
in 2020
• Network so complex that’s
Internet Map coloured by IP addresses
hard to draw on a map (Courtesy of W.R. Cheswick)
Prof. A. Liotta 2
3. Future Internet: what are we up against?
Explosive combination of sheer scale and diversity of things
Building
ICT
autom.
Security
Energy
& Safety
Consu-
mer & Retail
Home
Health- Transpor
care Indust. tation
Autom.
(*)
Source: Cisco data
forecast, Feb 2011
Prof. A. Liotta 3
4. What are we up against? Insurmountable barriers
Each ~100 mWatt transmission power Each 1 Gigawatt supply
Prof. A. Liotta 4
5. Sheer complexity (as in “complexity theory”)
• Properties of whole can’t be
inferred from properties of
individual parts
• Individual components
interact nonlinearly, leading
to emergent behavior
• Constantly evolves and
unfolds over time
Prof. A. Liotta 5
6. Natural networks work thanks to adaptivity and learning
Eminent examples:
• Human brain
• Immune system
• Ecosystems
• Insect colonies
• Most social structures
• …
Brain synapses
(Source: www.chiefscientist.gov.au)
Prof. A. Liotta 6
7. Today: a complex system operated simplistically
Over-dimensioned Over-dimensioned No idea whether
servers data flows user is satisfied
Over-provisioned Over-specified
networks terminals
7
8. Good news!
A billion dollars available to re-think the Internet
“The growth of the Internet is strictly
intertwined with socio-economic,
environmental and cultural developments”
EU FIRE program
Prof. A. Liotta 8
9. Unique opportunity to integrate knowledge
already available outside of the networking community
Phycology Law
Social
science
Physics Cybernetics
Natural Systems
sciences ecology
Computer Graph
science theory
Computational
intelligence
learn & evolve
FI taming complexity
Complexity
theory
Cognitive
Prof. A. Liotta Networks 9
10. Example: learning how to flip pancakes
Courtesy of: Petar Kormushev and Sylvain Calinon
Italian Institute of Technology (IIT)
Prof. A. Liotta 10
11. Networks too can learn how to deal with new
perturbations. Example video delivery conditions
News clip Sport clip
Original
Clips
50 50
100 100
150 150
Difference in 200 200
Temporal 250 250
300
300
Motion 350 350
400 400
450
450
500
500
550
550
100 200 300 400 500 600 700
100 200 300 400 500 600 700
Initially we train the Reinforcement Learning
network to handle teaches how to handle
Prof. A. Liotta “News over Laptop” “Sport over Phone” 11
12. Networks too can learn how to deal with new
perturbations. Example video delivery conditions
‘Sport over mobile phone’
QoS probe
actuators
Optimizing QoE Machine QoE measure
QoS prediction Learning or inference
Prof. A. Liotta 12
13. Can we use computational intelligence to build
‘nature-like’ computer networks?
Multi-disciplinarity
SOLUTIONS
Hidden patterns
Emergent behavior
Self-regulation
Learning
13
14. Thank you !
More about my work http://bit.ly/autonomic_networks
In the press http://bit.ly/press_articles
“All of YouTube through a 40-year-old funnel”
bit.ly/Volkskrant-EN bit.ly/booklet-antonio bit.ly/pervasive-networks
Prof. A. Liotta 14
Notes de l'éditeur
Major concern so far has been to provide network capacity and wireless access
Figure 4. Open-loop video delivery
Figure 9. Cognitive networks can learn how to differentiate ‘good’ from ‘bad’ signals, recognize viruses, and adjust to extreme perturbations.