The Public Option: A non-regulatory alternative to Network Neutrality

Transcription

1 The Public Option: A non-regulatory alternative to Network Neutrality Richard Ma Advanced Digital Sciences Center, Illinois at Singapore School of Computing, National University of Singapore Joint work with Vishal Misra (Columbia University)

2 The Internet Landscape Internet Service Providers (ISPs) Internet Content Providers (CPs) Regulatory Authorities Users/Consumers

3 Net Neutrality: Some History Early 2005, Madison River Communications Block VoIP $15,000 fine August 2008, Comcast Block Bittorrent packets The FCC imposed no fine, but required Comcast to end such blocking in the year April 6, 2010, Comcast Vs. FCC U.S. Court of Appeals ruled that the FCC has no powers to regulate any ISP.

4 Net Neutrality: Our Focus The content/application side of the twosided market. Classic example: night club Whether a neutral network is beneficial for end-users?

5

6 Network Neutrality (NN) Happy?

7 Paid Prioritization (PP) Happier?

8 Highlights A more realistic equilibrium model of content traffic, based on User demand for content System protocol/mechanism Game theoretic analysis on user utility under different ISP market structures: Monopoly, Duopoly & Oligopoly Regulatory implications for all scenarios and the notion of a Public Option

9 Three-party model (M, μ, N) λ i μ N M μ: capacity of a single access (eyeball) ISP M: # of users of the ISP (# of active users) N: set of all content providers (CPs) λ i : throughput rate of CP i N

10 User-side: 3 Demand Factors Unconstrained throughput θ i Upper-bound, achieved under unlimited capacity E.g. 5Mbps for Netflix Popularity of the content α i Google has a larger user base than other CPs. Demand function of the content D i (θ i ) Percentage of users still being active under the achievable throughput θ i θ i

11 Unconstrained Throughput λ i (Max) Throughput θ i (= 7Kbps) User size M(= 10) Content unconstrained throughput λ i = α i Mθ i (= 42Kbps) Content popularity α i (= 60%)

12 Demand Function D i θ i demanding # of users α i MD i θ i α i M θ i achievable throughput θ i

13 Demand Function D i θ i demanding # of users α i MD i θ i Assumption 1: D i θ i is continuous and non-decreasing in θ i with D i θ i = 1. More sensitive to throughput Throughput of CP i: α i M λ i θ i = α i MD i θ i θ i θ i achievable throughput θ i

14 System Side: Rate Allocation Rate allocation mechanism Θ d, μ maps fixed demands and capacity to throughput Axiom 1 (Throughput upper-bound) Θ i d, μ θ i Axiom 2 (Work-conserving or Pareto Opt.) λ N Θ d, μ = λ i Θ i d, μ i N = min μ, λ i i N

15 Rate Allocation Θ d, μ Axiom 3 (Consistency) There exists a family of continuous non-decreasing functions Θ γ = Θ i γ : i N such that Θ γ 1 Θ γ 2, γ 1 γ 2. For any d, μ, there exists a γ satisfying Θ d, μ = Θ γ

16 Uniqueness of Rate Equilibrium d, θ s. t. d = D θ θ = Θ d, μ θ = Θ D θ, μ Theorem (Uniqueness): A system (M, μ, N) has a unique equilibrium {θ i i N} (and therefore {λ i i N}) under Assumption 1 and Axiom 1, 2 and 3. User demand D i : θ i d i Rate allocation Θ: d i : i N, μ Rate equilibrium θ i, d i : i N θ i : i N

17 ISP Paid Prioritization ISP Payoff: c i P λ i = cλ P Capacity Charge Premium Class M, κμ, P κμ $c/unit traffic Ordinary Class M, 1 κ μ, O (1 κ)μ $0

18 Monopolistic Analysis Players: monopoly ISP I and the set of CPs N A Two-stage Game Model M, μ, N, I 1 st stage, ISP chooses s I = (κ, c) announces s I. 2 nd stage, CPs simultaneously choose service classes reach a joint decision s N = (O, P). Outcome (two subsystems): M, κμ, P : set P (of CPs) share capacity κμ M, 1 κ μ, O : set O share capacity 1 κ μ

19 Utilities (Surplus) ISP Surplus: IS = c i P λ i = cλ P ; Consumer Surplus: CS = i N φ i λ i φ i : per unit traffic value to the users Content Provider: v i : per unit traffic profit of CP i u i λ i = v i λ i v i c λ i if i O, if i P.

20 Type of Content Profitability of CP v i Value to users φ i

21 Monopolistic Analysis Players: monopoly ISP I and the set of CPs N A Two-stage Game Model M, μ, N, I 1 st stage, ISP chooses s I = (κ, c) announces s I. 2 nd stage, CPs simultaneously choose service classes reach a joint decision s N = (O, P). Theorem: Given a fixed charge c, strategy s I = (κ, c) is dominated by s I = (1, c). The monopoly ISP has incentive to allocate all capacity for the premium service class.

22 Utility Comparison: Φ vs Ψ ν = μ/m Φ = CS/M Ψ = IS/M

23 Regulatory Implications Ordinary service can be made damaged goods, which hurts the user utility. Implication: ISP should not be allowed to use non-work-conserving policies (κ cannot be too large). Should we allow the ISP to charge an arbitrarily high price c?

24 High price c is good when Profitability of CP v i Value to users φ i

25 High price c is bad when Profitability of CP v i Value to users φ i

26 Oligopolistic Analysis A Two-stage Game Model M, μ, N, I 1 st stage: for each ISP I I chooses s I = (κ I, c I ) simultanously. 2 nd stage: at each ISP I I, CPs choose service classes with s N I = (O I, P I ) Difference with monopolistic scenarios: Users move among ISPs until the per user utility Φ I is the same, which determines the market share of the ISPs ISPs try to maximize their market share.

27 Duopolistic Analysis P ISP I with s I = (κ, c) O N ISP J with s J = (0, 0)

28 Duopolistic Analysis: Results Theorem: In the duopolistic game, where an ISP J is a Public Option, i.e. s J = (0, 0), if s I maximizes the non-neutral ISP I s market share, s I also maximizes user utility. Regulatory implication for monopoly cases:

29 Oligopolistic Analysis: Results Theorem: Under any strategy profile s I, if s I is a best-response to s I that maximizes market share, then s I is an ε best-response for the per user utility Φ. The Nash equilibrium of market share is an ε-nash equilibrium of user utility. Oligopolistic scenarios:

30 Regulatory Preference ISP market structure Monopoly Oligopoly User Utility

31