ISP Settlement Problems: a New Angle via Network Economics

1. ISP Settlement Problems: a New Angle via Network Economics Tianbai Ma Department of Electrical Engineering Columbia University

2. Outline • Network Economics -- A new research area • My Current Research -- Highlight two problems • ISP profit sharing • ISP selfish behaviors • My Research -- Past and Future

3. Building blocks of the Internet: ISPs • The Internet is operated by hundreds of connected Internet Service Providers (ISPs). • An ISP is a business entity • Provides Internet services. • Common objective: to make profit.

4. Three types of ISPs • Eyeball (local) ISPs: • Provide Internet access to individual users. • E.g. PCCW • Content ISPs: • Server content providers and upload information. • Transit ISPs: • Tier 1 ISPs: global connectivity of the Internet. • Help other ISPs forward traffic.

5. A motivating example • Win-win under Client-Server: • Content providers find customers • Users get information • ISPs obtain revenue Fixed-rate charge Volume-based charge

6. What happens with P2P traffic? • Win-lose under P2P paradigm: • Content providers pay less • Customers get faster downloads • Content ISP obtains less revenue • Eyeball ISPs handle more traffic • In reality: P2P is a nightmare for ISPs.

7. Engineering solutions and … beyond • Proposed engineering approaches: • ISPs: Drop P2P packets based on port number • Users: Dynamic port selection • ISPs: Deep packet inspection • Users: Disguise by encryption • ISPs: Behavioral analysis • A new/global angle: Network Economics • Goal: a win-win/fair solution for all. • Related Issues: • What is a win-win/fair solution? How to find it? • How to design algorithms/protocols to achieve? • An exciting and upcoming area: • Inter-disciplinary：CS, Economics, Operation Research, Policy … • Stanford, CMU, UC Santa Cruz and etc. build new courses/department.

8. Two important issues of the Internet 1. Network Neutrality Debate: Content-based Service Differentiation ? Yes No Legal/regulatory policy for the Internet industry: Allow or Not? Allow: ISPs might dominate; Not allow: ISPs might die. Either way, suppress the development of the Internet. 2. Network Balkanization: Break-up of connected ISPs 15% of Internet unreachable Level 3 Cogent Not a technical/operation problem, but an economic issue of ISPs. Threatens the global connectivity of the Internet.

9. Two important issues of the Internet • A new/global angle: Network Economics • Goal: a win-win/fair solution for all. • Related Issues: • What is a win-win/fair solution? How to find it? • How to design algorithms/protocols to achieve?

10. Problems we try to solve • Problem 1: Find a win-win/fair profit-sharing solution for ISPs. • Challenges • What’s the solution? ISPs don’t know, even with the best intention. • How to find it? Complex ISPs structure, computational complexity. • How to implement? Need to be implementable for ISPs.

11. How to share profit? -- the baseline case • One content and one eyeball ISP • Profit V = total revenue = content-side + eyeball-side • Win-win/fair profit sharing:

12. Unique solution (Shapley value) How to share profit? – two symmetric eyeball ISPs Symmetry: same profit for symmetric eyeball ISPs Efficiency: summation of individual ISP profits equals v Fairness: same mutual contribution for any pair of ISPs Win-win/fair properties:

13. Myerson 1977 Efficiency Symmetry Fairness Shapley Value Shapley 1953 Efficiency Symmetry Dummy Additivity Young 1985 Efficiency Symmetry StrongMonotonicity History and properties of the Shapley value What is the Shapley value? – A measure of one’s contribution to different coalitions that it participates in.

14. How to share profit? -- n symmetric eyeball ISPs • Theorem: the Shapley profit sharing solution is

15. Results and implications of profit sharing • More eyeball ISPs, the content ISP gets larger profit share. • Users may choose different eyeball ISPs; however, must go through content ISP, • Multiple eyeball ISPs provide redundancy， • The single content ISP has leverage. • Content’s profit with one less eyeball: • The marginal profit loss of the content ISP: If an eyeball ISP leaves • The content ISP will lose 1/n2 of its profit. • If n=1, the content ISP will lose all its profit.

16. Profit share -- multiple eyeball and content ISPs • Theorem: the Shapley profit sharing solution is

17. Results and implications of ISP profit sharing • Each ISP’s profit share is • Inversely proportional to the number of ISPs of its own type. • Proportional to the number of ISPs of the opposite type. • Intuition • When more ISPs provide the same service, each of them obtains less bargaining power. • When fewer ISPs provide the same service, each of them becomes more important.

18. Profit share -- eyeball, transit and content ISPs • Theorem: the Shapley profit sharing solution is

19. Profit share -- Shapley value under general topologies 1. Shapley values under sub-topologies: Theorem: (Dynamic Programming) 2. Whether the profit can still be generated:

20. Current ISP Business Practices: A Macro Perspective Two forms of bilateral settlements: Zero-Dollar Peering Provider ISPs $$$ $$$ Customer-Provider Settlement Customer ISPs

21. $ $ $ $ $ $ Achieving the win-win/fair profit share $ $ $ Eyeball-side Revenue $ Content-side Revenue $ $ $ $ $ $ $ $

22. $ $ $ $ $ $ Achieving the win-win/fair profit share Eyeball-side Revenue $ Content-side Revenue $ $ $ $ $ • Two revenue flows to achieve the Shapley profit share: • Content-side revenue: Content Transit Eyeball • Eyeball-side revenue: Eyeball Transit Content

23. $ $ $ $ $ $ Achieving Shapley solution by bilateral settlements Eyeball-side Revenue $ Content-side Revenue Providers $ $ Customers Customers Zero-dollar Peering $ $ $ • When CR ≈ BR, bilateral implementations: • Customer-Provider settlements (Transit ISPs as providers) • Zero-dollar Peering settlements (between Transit ISPs) • Current settlements can achieve fair profit-share for ISPs.

24. $ $ $ $ $ $ Achieving Shapley solution by bilateral settlements $ $ $ $ $ $ $ $ $ Eyeball-side Revenue $ $ $ $ $ $ Customer Provider Paid Peering Content-side Revenue $ $ $ • If CR >> BR, bilateral implementations: • Reverse Customer-Provider (Transits compensate Eyeballs) • Paid Peering (Content-side compensates eyeball-side) • New settlements are needed to achieve fair profit-share.

25. Problems we try to solve • Problem 1: Find a win-win/fair profit-sharing solution for ISPs. • Challenges • What’s the solution? ISPs don’t know, even with the best intention. • Answer: The Shapley value. • How to find it? Complex ISPs structure, computationally expensive. • Result: Closed-form solution and Dynamic Programming. • How to implement? Need to be implementable for ISPs. • Implication: Two new bilateral settlements.

26. Recap: ISP Practices from a Macro Perspective Two current forms of bilateral settlements: Our Implication: Two new forms of bilateral settlements: Zero-Dollar Peering $$$ Inverse Customer-Provider Settlement Paid Peering $$$ Customer-Provider Settlement Next, a Micro Perspective Inspection about ISP behaviors

27. Problems we try to solve • Problem 1: Find a win-win/fair profit-sharing solution for ISPs. • Challenges • What’s the solution? ISPs don’t know, even with the best intention. • Answer: The Shapley value. • How to find it? Complex ISPs structure, computationally expensive. • Result: Closed-form solution and Dynamic Programming. • How to implement? Need to be implementable for ISPs. • Implication: Two new bilateral settlements. • Problem 2: Encourage ISPs to operate at an efficient/optimal point. • Challenges • How to induce good behavior? Selfish behavior/strategy hurt others. • What is the impact of the entire network? Equilibrium is not efficient.

28. Current ISP Business Practices: A Micro Perspective Provider ISP Three levels of ISP decisions • Interconnecting decision E • Routing decisions R (via BGP) • Bilateral settlements f Interconnection withdrawal Settlement faffects E, R Customer-Provider Settlements provider charges high Route change Hot-potato Routing Source Destination Zero-Dollar Peering Shortest Path Routing Customer ISP Customer ISP

29. W = 1 An ideal case of the ISP decisions Fixed Revenue Backbone ISP 1 Local ISP 1 Local ISP 2 A simple example: Peering links at both coasts Locally connect to both backbone ISPs Two backbone ISPs Two local ISPs End-to-end service generates revenue Routing costs on links Backbone ISP 2

30. W = 1 An ideal case of the ISP decisions Global Min Cost MIN routing cost and MAX profit Fixed Revenue Cost|Profit x22/4 x12/16 x32/16 1/2 A simple example: Two backbone ISPs Two local ISPs End-to-end service generates revenue Routing costs on links x42/8 x52/8 1/2 x62/16 x82/16 x72/4 We normalize the total required traffic load to be 1.

31. Problems with the current practice Global Min Cost Topology Balkanization Well-connected topology Increased routing and reduced profit MIN routing cost and MAX profit Cost|Profit x22/4 x12/16 x32/16 1/2 A simple example: Two backbone ISPs Two local ISPs End-to-end service generates revenue Routing costs on links x42/8 x52/8 1/2 x62/16 x82/16 x72/4 Behavior 1: ISPs interconnect selfishly to maximize profits! e.g. Backbone ISPs charge local ISPs.

32. Problems with the current practice Global Min Cost Hot Potato Topology Balkanization Increased routing and reduced profit Cost|Profit Profit reduction by routing inefficiency x22/4 x12/16 1/2 A simple example: Two backbone ISPs Two local ISPs End-to-end service generates revenue Routing costs on links x42/8 x52/8 1/2 1 x82/16 x72/4 Behavior 1: ISPs interconnect selfishly to maximize profits! Behavior 2: ISPs route selfishly to maximize profits! e.g. upper backbone ISP wants to use hot-potato routing to reduce its routing cost.

33. Issues of Current ISP Settlement: A Micro Perspective Global Ideal case: cooperative ISPs ISPs selfishly interconnect ISPs selfishly route traffic How to encourage ISPs to operate at efficient/optimal points?

34. $$$ j(E,R) $$ $$ Our solution: The Shapley Mechanism j Provider ISP Recall: three levels of ISP decisions • Interconnecting decision E • Routing decisions R • Bilateral settlements f Multilateral settlements j jcollects revenue from customers jdistributes profits to ISPs Settlement faffects E, R Customer-Provider Settlements Source Destination Zero-Dollar Peering Customer ISP Customer ISP

35. ISP behaviors under the Shapley mechanism j Given:j Local decisions:Ei,Ri Objective: to maximizeji(E,R) Ei Ri

36. Results: Routing Incentive Hot Potato Global Min Cost Local Min Cost j(E,R) Recall the inefficiency situation Cost|Profit x22/4 Shapley mechanism distributes profit x12/16 Profit maximized Profit increase 1/4 1/2 E.g. the upper ISP wants to minimize local routing cost x42/8 x52/8 1 3/4 1/2 Best strategy for all ISPs: global min cost routing x82/16 x72/4 ISPs route selfishly to maximize profits!

37. Results: Incentives for using Optimal Routes • Given any fixed interconnecting topology E, ISPs can locally decide routing strategies {Ri*} to maximize their profits. • Theorem (Incentive for routing): Any ISP i can maximize its profitjiby locally minimizing the global routing cost. • Implication: ISPs adapt to global min cost routes. • Corollary(Nash Equilibrium):Any global min cost routing decision is a Nash equilibrium for the set of all ISPs. • Implication: global min cost routes are stable. Surprising! Local selfish behaviors coincide with global optimal solution!

38. Results: Interconnecting Incentive Global Min Cost j(E,R) Recall: the best strategy for all ISPs is to use global min cost routes. Cost|Profit x22/4 Profit increase x12/16 x32/16 E.g. the left local ISP connects to the low backbone ISP. 5/12 1/2 x42/8 x52/8 1/2 7/12 Further the right local ISP connects to the upper backbone ISP. x62/16 x82/16 x72/4 Profit increase ISPs interconnect selfishly to maximize profits!

39. Results: Incentive for Interconnecting • For any topology, a global optimal route R* is used by all ISPs. ISPs can locally decide interconnecting strategies {Ei*} to maximize their profits. • Theorem (Incentive for interconnecting):After interconnecting, ISPs will have non-decreasing profits. • Implication: ISPs have incentive to interconnect. • Does not mean: All pairs of ISPs should be connected. • Redundant links might not reduce routing costs. • Sunk cost is not considered. The incentive results (routing and interconnecting) have attracted attentions, e.g. Jean Walrand (UC Berkeley), Ramesh Johari (Stanford) Peter Key (Microsoft Research), R. Srikant (UIUC) and etc.

40. Micro perspective ISP behaviors: result summary Selfishly interconnect and route j solves the selfish interconnecting problem ISPs have incentive to use optimal routes j j solves the selfish routing problem ISPs have incentive to interconnect j

41. Problems we try to solve • Problem 1: Find a win-win/fair profit-sharing solution for ISPs. • Challenges • What’s the solution? ISPs don’t know, even with the best intention. • Answer: The Shapley value • How to find it? Complex ISPs structure, computationally expensive. • Result: Closed-form sol. and Dynamic Programming • How to implement? Need to be implementable for ISPs. • Implication: Two new bilateral settlements • Problem 2: Encourage ISPs to operate at an efficient/optimal point. • Challenges • How to induce good behavior? Selfish behavior may hurt other ISPs. • Answer: The Shapley mechanism • Result: Incentives for optimal routes and interconnection • What is the impact of the entire network? Equilibrium is efficient? • Result: Optimal global Nash equilibrium

42. Myerson 1977 Efficiency Symmetry Fairness Shapley 1953 Efficiency Symmetry Dummy Additivity Young 1985 Efficiency Symmetry StrongMonotonicity New Application and Properties of the Shapley Value Shapley Value: Contributions to Coalitions Ma et al. 2008 Ma et al. 2007 ISP Routing Incentive ISP Profit Sharing ISP Interconnecting Incentive • Guideline for • ISPs: negotiate stable and incentive settlements. • Governments: make regulatory policy for the industry.

43. My research: current work

44. My research: past work

45. My research: future interests Modeling: Optimization, Stochastic Process Distributed Systems and Networks Implementation: Algorithm, Protocol Design, Stability and Convergence Analysis Algorithmic Game Theory Algorithmic Mechanism Design Network Economics Mechanism Design: e.g. Tax Policy, Auction Theory Non-cooperative Game Theory Micro-economics and Game Theory Coalition Game Theory

46. Interdisciplinary research • Problem 1: Find a win-win/fair profit-sharing solution for ISPs. • Challenges • What’s the solution? ISPs don’t know, even with the best intention. • Answer: The Shapley value (Coalition Game) • How to find it? Complex ISPs structure, computationally expensive. • Result: Closed-form sol. and Dynamic Programming (Computer Sci.) • How to implement? Need to be implementable for ISPs. • Implication: Two new bilateral settlements (Operations) • Problem 2: Encourage ISPs to operate at an efficient/optimal point. • Challenges • How to induce good behavior? Selfish behavior may hurt other ISPs. • Answer: The Shapley mechanism (Mechanism design) • Result 1: Incentives for optimal routes and interconnection (Behaviors) • What is the impact of the entire network? Equilibrium is efficient? • Result: Optimal global Nash equilibrium(Non-cooperative Game)

47. My research: future research problems • Economics-related problems: • Facilitate P2P traffic on the Internet • Network Neutrality • Networking problems: • IP multicast protocols • Wireless mesh networks • Problems in other fields • Graph theory, Approximation algorithm, Complexity • Distributed database system • E-commerce, social networks

49. v( ) =0.8125 D ( ) = v( ) = 0.625 D ( ) = v( ) - v( ) = 0.1875 v( ) = 0 v( ) = 0.625 v( ) - The Shapley value mechanism j Revenue Profit v(S) is defined on any subset of ISPs Profit = Revenue - Cost Routing cost Profit: v(S) x22/4 Marginal contribution of ISP i to set of ISPs S: Di(S). x12/16 x32/16 1/2 x42/8 x52/8 1 1/2 x62/16 x82/16 x72/4

50. S(p, ) j( )=2.4/6=0.4 p D(S(p, )) v( )=0 v( )=0 Empty Empty v( )- v( )- v( )=0.2 v( )=0.6 v( )=0.8 v( )=0.8 v( )- v( )- The Shapley value mechanism j N: total # of ISPs, e.g. N=3 P: set of N! orderings S(p,i): set of ISPs in front of ISP i