Is a MarketDriven Network Infrastructure Self Sustaining

1
Is a Market-Driven Network Infrastructure Self
Sustaining?

• Ramesh Johari
• August 29, 2005

2
Big picture

• Focus competition and cooperation between
  Internet service providers
• Key question
• When do competitive forces succeed in building
  network infrastructure?

3
Outline

• Background on ISP contracting
• Economic explanations fortodays problems
• The role of information
• Questions for the future

4
ISP contracts

• Internet 1000s of ASes (autonomous systems)
• Bilateral contracts between ASes
• Transit vs. peer contracts

A
B
5
Problems in the ISP industry

• In 2002, seven dominant players
• Sprint
• ATT
• MCI/UUnet
• Qwest
• CW
• Level3
• Genuity

6
Problems in the ISP industry

• In 2002, seven dominant players
• Sprint (subsidized by wireless)
• ATT ACQUIRED (SBC)
• MCI/UUnet ACQUIRED (Verizon)
• Qwest 18B debt
• CW R.I.P. (in U.S.)
• Level3 (merged w/Genuity)
• Genuity ACQUIRED (Level3)

7
Econ 101, pt. 1 fixed costs

• The Internet backbone has
• high fixed cost of startup
• low marginal cost of operation
• ) regulated monopoly is efficient

8
Econ 101, pt. 2 war of attrition

• Pricing below marginal cost
• ) War of attrition (repeated game)
• Lose money now in hopes of being last firm
  standing

9
Econ 101, pt. 3 Bertrand

• Example
• If p1

CNN
p1
p2
ISP 1
ISP 2
peer
eyeballs
10
Is Econ 101 enough?

• Econ 101 captures the essence
• cutthroat pricing
• massive financial losses
• last firm standing (t.b.d.)
• Question
• Is a regulated monopoly the only endgame?

11
Answer 1 Integration

• Vertical integration ) competition
• Idea
• Local service revenue subsidizes high cost
  investment

12
Answer 2 Access vs. core

• While core may be commoditized,low transit
  prices naturally encourageaccess
  competition(e.g., WiFi vs. Australian ISP
  market)

13
Answer 3 Engineering

• What is the problem in Bertrand example?
• ISP 2 receives no credit for the value generated.
• Current protocols dont expedite transmission of
  value information.

14
The role of information

• Providers dont understand value
• Examples
• Early peering negotiations
• U.S. vs. the world
• AOL vs. Akamai
• Informative packet routing protocols )
• value identification

15
Problem statement

• Is more information flow a good thing?
• Example
• Two providers, A and B
• Cost per unit flow, CA(fA) and CB(fB)
• Price per unit flow, pA and pB
• Good incentive price to avoid congestion
• Bad incentive profit maximization

16
Settlements

• Could a settlement-based scheme communicate
  value?
• Yes, but only if propagated to the ends.
• (Agree? Disagree?)

17
Future discussion

• Some aspects of the economic viability
  discussion are pure economics.
• But network engineering affects information flow,
  and this hasa first order impact on market
  structure.