Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Internet topology dynamics in ten minutes
Sergey Kirgizov under the supervision of Clémence Magnien
Complex Networks ⊂ LIP...
Outline

1 What do we observe?

2 Why it is so important?

3 How do we study our observations?

2
What do we observe?
Internet IP-level topology

Nodes:
Links:

IP addresses
connections between hosts

3
Internet IP-level topology

Impossible to obtain a full map

Nodes:
Links:

IP addresses
connections between hosts

3
Ego-centered views

d1

d2

m

(shortest) paths between monitor and destinations

4
Ego-centered views

d1

d2

m

a measurement by tracetree

4
Ego-centered views

d1

d2

m

another measurement by tracetree
load-balancing
4
Ego-centered views

d1

d2

m

yet another measurement by tracetree
evolution of routes
4
Ego-centered view dynamics

d2

d1

m

d1

d2

m

d1

d2

m

Time

Fast periodic measurements =⇒ study of the dynamics

5
Why we should study this?
Some possible applications

• Develop a good model of the network
• Security (unstable routes means more spyware?)
• Robus...
Can we see the dynamics?
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
Evolution of symmetric difference between
measurements (one destination)
Measurement is a set of IP-links
m1 : first measure...
1 min 30 sec

3 min

1

0 2 4 6 8

Delay:

0

500
Destinations

1000

2000

3000

0

1000

2000

3000
1 min 30 sec

3 min

1

0 2 4 6 8

Delay:

2000

3000

0

1000

2000

3000

0

1000

2000

3000

0

1000

2000

3000

400 ...
Overview of problems and methods

number of
destinations

?
frequency
of observations

observed
evolution

Methods:
• Real...
Questions?
http://kirgizov.complexnetworks.fr/
Upcoming SlideShare
Loading in …5
×

Internet topology dynamics in 10 minutes

283 views

Published on

Masterclass-discussion with Vint Cerf.

A 10 minutes of introduction into Internet Topology Dynaimcs.
In this presentation, I answer the following questions:

1. What is the Internet Topology Dynamics?
2. Why it is so important to study such dynamics?
3. How can we _see_ the dynaimcs?
4. Which methods we use to study the dynaimcs?

Published in: Education
  • Be the first to comment

  • Be the first to like this

Internet topology dynamics in 10 minutes

  1. 1. Internet topology dynamics in ten minutes Sergey Kirgizov under the supervision of Clémence Magnien Complex Networks ⊂ LIP6 ⊂ (UPMC × CNRS) 4 March 2014
  2. 2. Outline 1 What do we observe? 2 Why it is so important? 3 How do we study our observations? 2
  3. 3. What do we observe?
  4. 4. Internet IP-level topology Nodes: Links: IP addresses connections between hosts 3
  5. 5. Internet IP-level topology Impossible to obtain a full map Nodes: Links: IP addresses connections between hosts 3
  6. 6. Ego-centered views d1 d2 m (shortest) paths between monitor and destinations 4
  7. 7. Ego-centered views d1 d2 m a measurement by tracetree 4
  8. 8. Ego-centered views d1 d2 m another measurement by tracetree load-balancing 4
  9. 9. Ego-centered views d1 d2 m yet another measurement by tracetree evolution of routes 4
  10. 10. Ego-centered view dynamics d2 d1 m d1 d2 m d1 d2 m Time Fast periodic measurements =⇒ study of the dynamics 5
  11. 11. Why we should study this?
  12. 12. Some possible applications • Develop a good model of the network • Security (unstable routes means more spyware?) • Robustness of the network and protocols • Event detection • Web caching • etc 6
  13. 13. Can we see the dynamics?
  14. 14. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  15. 15. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  16. 16. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  17. 17. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  18. 18. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  19. 19. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  20. 20. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  21. 21. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  22. 22. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  23. 23. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  24. 24. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  25. 25. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  26. 26. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  27. 27. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  28. 28. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 2 4 6 8 i 10 12 14 (delay ≈ 1 min 30 sec) 7
  29. 29. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 5 10 i 15 20 (delay ≈ 1 min 30 sec) 7
  30. 30. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 0 10 20 30 i 40 50 (delay ≈ 1 min 30 sec) 7
  31. 31. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 0 20 40 60 i 80 100 (delay ≈ 1 min 30 sec) 7
  32. 32. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 5 10 15 20 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 0 200 400 600 i 800 1000 (delay ≈ 1 min 30 sec) 7
  33. 33. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 10 15 20 5 Load-balancing 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 0 200 400 600 i 800 1000 (delay ≈ 1 min 30 sec) 7
  34. 34. Evolution of symmetric difference between measurements (one destination) Measurement is a set of IP-links m1 : first measurement mi : i-th measurement m1 mi 10 15 20 5 Evolution 0 d(m1 , mi ) d(m1 , mi ) = |m1 ∩ mi | 0 50000 150000 i 250000 (delay ≈ 1 min 30 sec) 7
  35. 35. 1 min 30 sec 3 min 1 0 2 4 6 8 Delay: 0 500 Destinations 1000 2000 3000 0 1000 2000 3000
  36. 36. 1 min 30 sec 3 min 1 0 2 4 6 8 Delay: 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 400 800 1000 0 500 0 Destinations
  37. 37. Overview of problems and methods number of destinations ? frequency of observations observed evolution Methods: • Real-world measurements • Simulated measurements using random graphs • Theoretical study of dynamic random graphs • Stochastic process estimation from partial observations 9
  38. 38. Questions? http://kirgizov.complexnetworks.fr/

×