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

2. Network Economics -- A new research area
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: Fixed-rate charge
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
Let’s see two important issues of the Internet. Network Neutrality is an on-going and heated debate in both academia and the Internet industry. It argues whether content-based service differentiations should be provided for the Internet. Transit and eyeball ISPs support to provide differentiated services in order to generate more revenue. Content providers and users advocate the neutrality of the network which prohibits dropping packets based on their content, e.g. P2P packets in our previous example. This debate is very important, because it will influence the legal and regulatory policies for the Internet. If the government allows content-based service differentiation, ISPs might be able to charge at a high price and inhibit the development of companies that use the Internet, e.g. the content providers. However, if the government prohibits content discrimination, ISPs might not be able to generate enough revenue to cover their cost, and therefore cannot survive. In either way, it is going to suppress the development of the Internet. Another issue is Network Balkanization. Balkanization means the break-up of previously connected ISPs. In 2005, a transit ISP, Level 3, disconnected the links with another ISP, Cogent, resulting at least 15% of the Internet unreachable for the customers of both ISPs at that time. This was not a technical or operation problem; but rather an economic issue of the ISPs. Because Cogent refused to pay the price charged by Level 3, Level 3 disconnected the links. The consequence is that it threatens the global connectivity of the Internet.
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.
In particular, we target on the problem of finding a win-win/fair profit-sharing solution for ISPs. The first challenge is that ISPs, even with the best intention to cooperate, do not know what a win-win/fair solution is. Second, even the ISPs have some idea of a fair solution; it might be difficult for them to find it, because of complex ISP structures as well as the computational complexity of calculating it. And we also need to see if the solution can be implemented in practice.

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

13. 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.
Shapley Value
Efficiency
Symmetry
Fairness
Myerson 1977
Dummy
Additivity
Shapley 1953
Strong Monotonicity
Young 1985
Here we briefly introduce the history and properties of the Shapley value. Shapley value models a cooperative scenario where sets of players can generate profits. The Shapley value of each player is a function of that player’s contribution to different coalitions that it participates in. Lloyd Shapley first characterized the solution by Efficiency, Symmetry, Dummy and Additivity properties. Later, Myerson used Fairness property to characterize the value. Young showed that the value can also be characterized by a Strong Monotonicity property. Notice, the Shapley value satisfies all six properties and the subsets of these properties uniquely determine the Shapley value solution.

14. How to share profit? -- n symmetric eyeball ISPs
Let’s continue our discussion. We can extend the result to the case where we have n eyeball ISPs serving the region. To satisfy the desirable properties, we uniquely determine the profit sharing solution for ISPs as the following.
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.
Let’s get a closer look at this solution. It suggests that the more of eyeball ISPs, the larger profit for the content ISP. Intuitively, this is because when an eyeball leaves, its customer can always go to another eyeball, and there is no revenue loss. So multiple eyeballs provide redundancy and the only content ISP can leverage. Using the formula, we can derive the fair share of the content ISP with one less eyeball ISP in the system. Then we can measure the marginal profit loss of the content ISP contributed by an eyeball ISP. The result says when an eyeball leaves the system, it will only cost one over n squared of the content’s original profit. In particular, if n=1, when the only eyeball leaves the system, it will cost the content ISP all its original profit. I am not sure if the following analogy is appropriate, but just imagine the only content ISP as Windows and the multiple eyeball ISPs as compatible processor-producer. The result explains why Microsoft naturally dominates the market and obtains high profit margin.

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.
These results are very interesting because they have the following implications. This solution says that each ISP’s profit share is inversely proportional to the number of ISPs of the same type, and is proportional to the number of ISPs of the opposite type. Intuitively, when more ISPs provide the same service, each of them becomes less important, and therefore gets less profit-share. When fewer ISPs provide the same service, each of them becomes more important, and therefore gets larger profit-share. This solution generalizes the previous result where m=1.
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
After getting the theoretic profit share for ISPs, let’s explore how we can implement the solution in practice. Let’s first look at the current practices of the ISPs. From a Marco-perspective, ISPs often maintain two forms of bilateral settlements. We illustrate them using following example, where information goes from the content side, traversing two transit ISPs and reaching an eyeball ISP. The first form is the customer/provider settlement which often involves a transit ISP playing the role of a provider to help other ISPs forward traffic and get paid from them. The second form is called zero-dollar peering, under which ISPs get connected, forward each other’s traffic, but do not pay each other.
Customer ISPs

21. Achieving the win-win/fair profit share$ $ $
Eyeball-side Revenue $
Content-side Revenue $ $ $ $ $ $ $ $ $ $
After getting the theoretic profit share, let’s explore its implications for ISP settlements. We focus on the three groups of ISPs. We know that Eyeball and Content ISPs receive end payments from customers and content providers. They keep their fair share of the revenue and forward the remaining. The Transit ISPs receive revenues from both sides and keep their own portion and keep forwarding the remaining revenue to the opposite side.

22. Achieving the win-win/fair profit share
Eyeball-side Revenue $
Content-side Revenue $ $ $ $ $ $ $
We can clearly see these two revenue flows.
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 $ $ $
Now, we consider how we can achieve the same solution using bilateral settlements. We first consider the case where the amount of content-side revenue is close to that of the eyeball-side as illustrated here. We can simply make the net money exchange between the Transit ISPs and the other two groups of ISPs. The resulting bilateral settlements are in the form of customer/provider relationship, where the Transit ISPs are the providers, and the zero-dollar peering between transit ISPs, where they keep their own revenues. This explains that when local ISPs were not so different from each other, these two kinds of settlements were stable, because they achieved a solution which is close to the Shapley value solution.
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 $ $ $
However, with the fast business development of the Internet, the content-side revenue is considerably much more than the fixed payments from eyeball side as illustrated here. After making the net money exchange again, the resulting bilateral settlements show that Transit ISPs need to compensate the Eyeball ISPs, which creates an inverse customer/provider relationship. In addition, since the payments from both sides are imbalanced, content-side ISPs need to compensate eyeball-side ISPs, which creates a paid-peering relationship. Although these new bilateral settlements largely deviate from the current settlements, we envision them to emerge so as to maintain stable settlements among ISPs.
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
$$$
Inverse Customer-Provider Settlement
Paid Peering
Customer-Provider Settlement
$$$
From a Marco-perspective, ISPs often maintain the following two bilateral settlements. Under the zero-dollar peering, Transit ISPs forward each other’s traffic without payment. The customer/provider settlement often involves a transit ISP playing the role of a provider to help other ISPs forward traffic and get paid from them.
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 f affects E, R
Customer-Provider
Settlements
provider charges high
Hot-potato Routing
Route change
Source
From now on, we consider more generic ISPs, which might have customers of individual users, corporations or a combination of them. From a Micro-perspective, ISPs make three levels of decisions to maximize their profits. First, ISPs make interconnecting decisions. They decide whether or not to connect with a neighboring ISP. We denote the interconnecting decisions as E, which can be considered as the set of links of the topology, because these decisions form the network topology. After the ISPs are connected, they make routing decisions. They exchange BGP advertising routes and based on their own preferences to choose routing paths. For example, an ISP has some traffic to send to a destination. It can choose a shortest path going through the two routers of its own, or it can use hot-potato routing, choosing the nearest egress point, and make other ISPs forward the traffic. Because ISPs are selfish, the hot-potato routing is more common in practice. On top of routing and interconnecting, ISPs make bilateral settlements as mentioned before. Because the ultimate objective of each ISP is to maximize its profit, the settlements affect the underlying routing and interconnecting decisions. For example, due to the service charge, a customer-ISP might withdraw the connection with a provider ISP, and as a result, it has to route through its peering link.
Destination
Zero-Dollar
Peering
Shortest Path Routing
Customer ISP
Customer ISP

29. An ideal case of the ISP decisions
Route
Topology
An ideal case of the ISP decisions
Fixed Revenue
Backbone ISP 1
W = 1
Local ISP 1
Local ISP 2
A simple example:
Locally connect to
both backbone ISPs
Peering links at
both coasts
Two backbone ISPs
Two local ISPs
End-to-end service generates revenue
Routing costs on links
Here, we illustrate an ideal case of the ISPs’ decisions. Consider we have two backbone ISPs covering the whole United States. They have routers on both the east and the west coast. Both ISPs peer with each other at both sites. Suppose we also have two local ISPs, one on each side. They interconnect with both backbone ISPs. Now, we have a well-connected network. Assume the customers of these local ISPs want to communicate, and the service can generate a certain amount of revenue. We normalize the revenue to be one, which is represented by the green bar. Although we ignored the costs in the Macro-perspective models, we consider the routing cost here.
Backbone ISP 2

30. An ideal case of the ISP decisions
Route
Topology
An ideal case of the ISP decisions
Global Min Cost
MIN routing cost and MAX profit
Fixed Revenue
Cost | Profit
x22/4
W = 1
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
We normalize total required traffic load for the service to be one and we assume the following cost functions in this example. Now what we are concerned about is the total profit that can be generated. Ideally the global min-cost route minimizes the aggregate routing cost. This route evenly splits the traffic on these two paths. The red bars represent the routing costs on each link. For example, this cost is one-sixteenth, which is 1/2 squared divided by four. The aggregate routing cost is shown by the red bar and the green bar represents the profit all these ISPs can obtain. Notice that the total length of the bar is the constant revenue from the service. So ISPs can cooperatively maximize the total profit.
x62/16
x82/16
x72/4
We normalize the total required traffic load to be 1.

31. Problems with the current practice
Route
Topology
Global Min Cost
Well-connected topology
Topology Balkanization
MIN routing cost and MAX profit
Increased routing and reduced 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
However, the reality is not so perfect. Selfish ISPs want to maximize their own profits by making interconnecting decisions. For example, because backbone-ISPs normally charge local ISPs for the transit services, local ISPs might not want to connect to all of them. The topology changes from a well-connected graph to be a barely connected graph. After adapting to the new topology, the achievable min-cost route becomes this. Accordingly, the routing cost increases and the profit for all the ISPs decreases. Notice that although financial transactions happen between ISPs, the aggregate profit for them is a constant sum represented by the green bar.
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
Route
Topology
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
Further, ISPs will maximize their own profits by choosing routing decisions. For example, the green backbone ISP can use hot-potato routing and shift all the traffic going through its own network to the peering link, making the other backbone ISP forward the traffic across the country. We can see that the routing cost for the other backbone ISP increases dramatically. As a result, this further increases the routing cost and reduces total profit.
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
Route
Topology
Cost | Profit (maximized)
Connectivity
Global Min Cost
Well-connected
ISPs selfishly interconnect
Route
Topology
Cost | Profit (reduced)
Connectivity
Global Min Cost
Balkanized
ISPs selfishly route traffic
From a Micro-perspective, we summarize the issues as follows. Ideally, we can have a well-connected network, where ISPs fully cooperate to minimize routing costs and maximize the aggregate profit. Due to the selfish interconnecting behaviors, the topology starts to Balkanize and the total profit reduces. The connectivity is not as robust as before. Further, due to the selfish routing behaviors, the routes used by the ISPs do not minimize global routing cost. Consequently, it further reduces the aggregate profit. As a result, the Internet is not as efficient as it could be, and the selfish behaviors of the ISPs hurt the profits for themselves.
Route
Topology
Cost | Profit (further reduced)
Connectivity
Hot Potato
Balkanized
How to encourage ISPs to operate at efficient/optimal points?

34. 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
j collects revenue from customers
j distributes profits to ISPs
Settlement f affects E, R
j(E,R)
$$$ $$
Customer-Provider
Settlements
Source
Recall the three levels of ISP decisions. Because the financial settlement affects the underlying routing and interconnecting decisions, we try to solve the selfish routing and interconnecting problems by re-designing the financial settlement. We replace the current bilateral settlements with a multilateral settlement. It collects revenue from all customers and distribute to individual ISPs. We denote phi as the profit distribution mechanism, which is a function of both the interconnecting topology and the routing decisions. Under this new mechanism, ISPs make underlying interconnecting and routing decisions to maximize their profits.
Destination
Zero-Dollar
Peering
Customer ISP
Customer ISP

35. ISP behaviors under the Shapley mechanism j
Given: j
Local decisions: Ei,Ri
Objective: to maximize ji(E,R)
From each ISP’s point of view, it knows this mechanism before it makes local interconnecting decisions Ei and local routing decisions Ri to maximize its profit phi(i). Ei Ri

36. Results: Routing Incentive
Route
Topology
Local Min Cost
Hot Potato
Global Min Cost
j(E,R)
Recall the inefficiency situation
Cost | Profit
x22/4
Shapley mechanism distributes profit
x12/16
Profit maximized
Profit increase
1/2
1/4
E.g. the upper ISP wants to minimize local routing cost
x42/8
x52/8 1
3/4
1/2
Recall the previous inefficient situation where the selfish backbone ISP uses hot-potato routing. Under our new settlement, the Shapley mechanism divides the profit to each individual ISP. Here, each color represents the profit of the ISP with the same color. Now, each ISP would consider to perform different routing strategies to maximize its own profit. For example, this backbone ISP might try to minimize the local routing costs going through its internal link and two peering links. As a result, it reduces global routing cost. After redistributing the profit, the backbone ISP’s profit is increased. Eventually, all ISPs will find that, if they perform a global min cost route, their individual profits will be maximized.
x82/16
Best strategy for all ISPs: global min cost routing
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 profit ji by 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.
Here, we formally state the result of incentive for using optimal routes. Assume we are given any fixed topology, we mean ISPs don’t change interconnecting decisions. Each ISP can locally choose the routes to maximize profit. Our theorem says that any ISP can maximize its profit under our profit distribution mechanism, if it uses a local route that minimizes the global routing cost. The theorem implies that whenever the current route is not optimal, each ISP has an incentive to change routes to be global optimal in order to maximize its own profit. A corollary from the theorem says, any global optimal route, which minimizes the global routing cost, is also a Nash equilibrium for all ISPs. The corollary implies that whenever the global optimal route is reached, the route is stable, because no ISP can obtain larger profit by unilaterally deviating from the optimal route. This result is quite important and surprising in the sense that the local selfish behavior coincides with the global optimal solution. In general it cannot be true. Without the mechanism, selfish behavior encourages hot-potato routing. The equilibrium is not efficient. The Shapley mechanism pulls the inefficient equilibrium to a global optimal equilibrium for all ISPs.
Surprising! Local selfish behaviors coincide with global optimal solution!

38. Results: Interconnecting Incentive
Route
Topology
Global Min Cost
Recall: the best strategy for all ISPs is to use global min cost routes.
j(E,R)
Cost | Profit
x22/4
Profit increase
x12/16
x32/16
1/2
5/12
E.g. the left local ISP connects to the low backbone ISP.
x42/8
x52/8
7/12
1/2
Further the right local ISP connects to the upper backbone ISP.
Let’s continue our example. Now we know, ISPs always have incentives to perform a global optimal route. Here, we assume they adapt to an optimal route whenever the topology changes. Now, ISPs start to think about using the interconnecting decisions to maximize their profits. For example, the local ISP tries to connect with the blue backbone ISP. After adapting to an optimal route, the cost reduces, and after the redistribution of profit, the profits of both connecting ISPs are increased. Similarly, the orange local ISP can also try to connect with the green backbone ISP. Now, the routing cost is minimized, and after redistributing the profit, both connecting ISPs’ profits are increased.
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.
We formalize the interconnecting incentive result here. We assume that given any topology, an optimal route will be used by all ISPs. Now ISPs can make local interconnecting decisions to change the topology. Our theorem says, whenever two ISPs interconnect, their profits are not going to decrease. Most of time, their profits are going to increase, because new link reduces global routing costs. This result implies that ISPs have incentive to interconnect under the Shapley value mechanism. However, it doesn’t mean all pairs of ISPs should be connected, because redundant links might not reduce routing costs, and the sunk cost for establishing a link 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
Route
Topology
Cost | Profit
Connectivity
Hot Potato
Balkanized
j solves the selfish interconnecting problem
ISPs have incentive to use optimal routes j
Route
Topology
Cost | Profit
Connectivity
Global Min Cost
Balkanized
A summary of the Micro-perspective results is the following. Current ISP practices show the selfish routing and interconnecting problems, which reduce the total profit and balkanize the topology. Under the Shapley value profit distribution mechanism, individual ISPs have incentives to use an optimal route to maximize profit for any given topology. Moreover, ISPs also have incentives to interconnect with each other to further increase their own profits. As a result, the aggregate profit can be maximized and the topology can be more robust and well-connected.
j solves the selfish routing problem
ISPs have incentive to interconnect j
Route
Topology
Cost | Profit
Connectivity
Global Min Cost
well-connected

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
Intuitively, since the Shapley value is a win-win/fair share solution, when individuals do good things to the whole community to increase social profit, its only fair share should be increased proportionally.

42. New Application and Properties of the Shapley Value
Shapley Value: Contributions to Coalitions
Efficiency
Symmetry
Dummy
Additivity
Shapley 1953
Efficiency
Symmetry
Fairness
Myerson 1977
Efficiency
Symmetry
Strong Monotonicity
Young 1985
Ma et al. 2008
Static properties are enriched by have dynamic properties …
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)
In particular, we target on a fundamental problem, which is try to find a win-win/fair profit sharing solutions for ISPs.
The first challenge is that, most of time, even with the best cooperative intentions, ISPs do not know what is a win-win/fair solution for them. Second, even the ISPs have some idea about a fair solution, sometimes it’s hard to find it, because the complex structures of the ISPs as well as computational complexity of the solutions.

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. The Shapley value mechanism j
Revenue
Profit v(S) is defined on any subset of ISPs
Profit = Revenue - Cost
Routing cost
Profit: v(S)
v( ) =0.8125
v( ) = 0.625
v( ) = 0
x22/4
Marginal contribution of ISP i to set of ISPs S: Di(S).
x12/16
x32/16
D ( ) =
v( ) -
v( ) =
1/2
Again, we consider to use the Shapley value to distribute profits. The profit function v is defined on each set of ISPs S. As we mentioned before, profit equals revenue minus cost. Suppose the set of ISPs can generate a fixed amount of revenue, they can perform a certain route with a fixed cost, then v(S) is the profit they generate. Using the previous example, if the four ISPs cooperate and use a min-cost route, the profit is maximized. Similarly, a subset of the ISPs can obtain lower profit due to a higher routing cost. If the set of ISPs are disconnected and cannot provide the service, the profit becomes zero. We further denote Delta_i(S) as the marginal contribution of ISP i to a set of ISPs S. For example, the marginal contribution of the green ISP to the set of remaining three ISPs is the profit difference between whether the green ISP is cooperating or not.
x42/8
x52/8
v( ) -
v( ) = 0.625
D ( ) =
1/2 1
x62/16
x82/16
x72/4

50. 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 p
Empty
S(p, )
D (S(p, ))
v( )=0
v( )=0.2
v( )-
v( )-
v( )=0.8
v( )=0.6
j( )=2.4/6=0.4
Then the Shapley value of an ISP can be calculated using the marginal contributions of this ISP to sets of other ISPs. N is the total number of ISPs in the system. Using an illustration of three ISPs, the summation is over the big pi, which is the set of N! orderings of the ISPs. Here we have the six different orderings of three ISPs. S(p,i) is the set of ISPs that is in front of I in the ordering pi. For example, if ISP i is the red ISP, these are the set S(pi,i). Now we can have the marginal contributions of the red ISP to the sets of ISPs. The Shapley value is the average of these N! marginal contributions.