Top Banner
1
21

Constantine Dovrolis Georgia Institute of Technology constantine@gatech

Feb 22, 2016

Download

Documents

magar

Constantine Dovrolis Georgia Institute of Technology [email protected]. Outline. What is this workshop about? The evolution of research ideas and industry transformations that led to this workshop Workshop objectives. What is this workshop about?. Background. Tier-1 network. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

1

Page 2: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

2

Constantine DovrolisGeorgia Institute of Technology

[email protected]

Page 3: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

3

Outline• What is this workshop about?

• The evolution of research ideas and industry transformations that led to this workshop

• Workshop objectives

Page 4: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

4

Enterprise customer

Transit Provider

Transit Provider

Enterprise customer

Content Provider

Content Provider

Tier-1 network

Tier-1 network

Tier-1 network

What is this workshop about?

$$

$$

$$

$$ $$

Page 5: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

5

What is this workshop about?

• Internet as a network of networks (Autonomous Systems)– Subject of Networking research

• Internet as a graph (AS-level graph)– Subject of Graph Theory research

• Internet as a networked market– Subject of Economics research

Page 6: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

6

But this workshop is also about• Real businesses, real money..– Transit providers (Level3, Cogent, ..)– Access providers (Comcast, Verizon, ..)– Content providers (Google, Yandex, ..)– Content Distribution Networks (Akamai, ..)– Internet Exchange Points (Equinix, ..)

• Regulatory bodies (e.g., FCC) and some hard policy questions – See Network Neutrality debate..

• Real people and their Internet experience

Page 7: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

7

The evolution of ideas/events that led to this workshop:

• Disclaimer: This is my own, biased view of this evolution

Page 8: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

8

The prehistoric era of the Internet(till about 20 years ago..)

Page 9: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

9

Topology modeling in the pre-historic era

Page 10: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

10

Followed by the Internet commercialization transition in 1995

Page 11: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

11

Early attempts to model Internet topology realistically (1996-7)

Page 12: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

12

At about the same time, “Internet economics” started attracting attention

Connectivity in the commercial InternetJ Cremer, P Rey, J Tirole The Journal of Industrial Economics, 2000

Page 13: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

13

A breakthrough in Internet topology modeling took place in 1999

Page 14: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

14

Followed by a rich literature in graph theory and network science about scale-free graphs

Page 15: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

15

But most of that work does not consider the various economic (and other) objectives that

drive Internet connectivity

Until the following paper in 2002..

Page 16: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

This breakthrough created a strong connection between Internet topology models, economics

and network formation games

16

Page 17: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

At about the same time, the networking community started measuring and modeling the

Internet as an evolving economic ecosystem

17

Page 18: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

And while the research community was blending together networking, economics and graph theory,

the industry was going through MAJOR transformations..

18

• Internet content originating mostly from CDNs• Global penetration of Internet Exchange Points

– Reducing peering costs and simplifying interconnections• Internet densification and “flattening”• Massive drop in Internet transit prices (price war?)• Consolidation of Internet transit providers• Major disputes between Tier-1 transit providers (e.g.,

Level3 vs Cogent in 2005)• Major disputes between Access and Content

providers/CDNs (e.g., Level3 vs Comcast in 2010)• Network neutrality debate: should ISPs charge

differently depending on content? Should they charge content providers?

Page 19: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

Bill Norton’s book/papers describe these transformations very clearly

19

Page 20: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

20

What does this all mean?Workshop Objectives

• The ground is ready for some truly cross-disciplinary work in the area of Internet Economics– Bring together Networking, Economics, Network

Science

• Real potential for major impact that can affect– Two-three billions of users?– Thousands of Internet-related companies?

• Plus, some new and interesting problems in each of the previous disciplines– Fertile ground to develop completely new

methods?

Page 21: Constantine  Dovrolis Georgia Institute of Technology constantine@gatech

21

A final note: How to avoid the “confusion of languages”?

• Our languages are very different:– Economics, networking,

theoretical CS, graph theory

– Can we communicate somehow?

– If we manage to communicate, the result can be larger than the sum of its parts