SLIDE 1
Subsidization Competition: Vitalizing the Neutral Internet
Richard T. B. Ma
School of Computing National University of Singapore
WIE 2014
SLIDE 2 Internet’s two-sided market
Problem is not in the transit market
Fiber optics backbone, rare congestion Competitive market with declining prices CPs bypass Tier-1 ISPs to improve performance
But in the mobile access market
High mobile infrastructure costs One-side pricing from end-users Lower profit margin than those of the CPs Few incentives for investments
SLIDE 3 About this work
Propose and study “subsidization competition”
CPs could voluntarily subsidize its users’ usage costs
Differences to sponsored data plan/”zero rate”
1.
Partial subsidization is allowed
2.
ISPs charge the same per-unit rate, regardless the source of revenue (no secret deals with CPs)
SLIDE 4 Basic system model 𝒏, 𝜈
Focus on an access ISP with capacity 𝜈 and
a set 𝒪 of CPs. For each 𝑗 ∈ 𝒪, denote
𝑛𝑗: user size, 𝜇𝑗: avg per user throughput 𝜄𝑗 ≜ 𝑛𝑗𝜇𝑗 as throughput and 𝜄 ≜
𝜄𝑗
𝑗∈𝒪
Define 𝜚 ≜ Φ 𝜄, 𝜈 as the system utilization
Φ 𝜄, 𝜈 ↗ 𝜄; Φ 𝜄, 𝜈 ↘ 𝜈 can be seen as system congestion
User throughput satisfies 𝜇𝑗 ≜ 𝜇𝑗 𝜚 ↘ 𝜚
SLIDE 5 Basic system model 𝒏, 𝜈
𝜚 is the utilization of a system 𝒏, 𝜈 iff
𝜚 = Φ 𝑛𝑗𝜇𝑗 𝜚
𝑗∈𝒪
, 𝜈
utilization is unique throughput of CPs
SLIDE 6
One-sided pricing model
If ISP charges 𝑞, its revenue is 𝑆 ≜ 𝑞𝜄 User size: 𝑛𝑗 ≜ 𝑛𝑗 𝑞 ↘ 𝑞
SLIDE 7 One-sided pricing model
Price effect:
𝜖𝜚 𝜖𝑞 ≤ 0; 𝜖𝜄 𝜖𝑞 ≤ 0.
CP 𝑗’s throughput 𝜄𝑗 increases with price 𝑞 iff
𝜗𝑞
𝑛𝑗/𝜗𝜚 𝜇𝑗 < −𝜗𝑞 𝜚
where 𝜗𝑦
𝑧 ≜ 𝜖𝑧 𝜖𝑦 𝑦 𝑧 denotes the x-elasticity of y.
𝜗𝑞 𝑛𝑗 small: users are not price sensitive 𝜗𝜚 𝜇𝑗 large: traffic is very sensitive to congestion
SLIDE 8
Subsidization model
Denote 𝑟 as a policy that limits the subsidy,
each CP 𝑗 choose to subsidize 𝑡𝑗 ∈ 0, 𝑟
Denote 𝒕 as the strategy profile of the CPs User size becomes 𝑛𝑗 = 𝑛𝑗 𝑢𝑗 = 𝑛𝑗 𝑞 − 𝑡𝑗 CP’s utility becomes 𝑉𝑗 = 𝑤𝑗 − 𝑡𝑗 𝜄𝑗 Define social welfare 𝑋 =
𝑤𝑗𝜄𝑗
𝑗∈𝒪
SLIDE 9
Subsidization model
SLIDE 10
Nash equilibrium
For price 𝑞 and policy 𝑟, a strategy profile 𝒕 is a
Nash equilibrium iff each 𝑡𝑗 solves
𝑁𝑏𝑦 𝑉𝑗 𝑡𝑗; 𝒕−𝑗 = 𝑤𝑗 − 𝑡𝑗 𝜄𝑗 𝒕 𝑡. 𝑢. 0 ≤ 𝑡𝑗 ≤ 𝑟.
There exists a unique Nash equilibrium if for
any 𝑡′ ≠ 𝑡, there always exist CP 𝑗 such that
𝑡𝑗
′ − 𝑡𝑗
𝑣𝑗 𝒕′ − 𝑣𝑗 𝒕 < 0
where 𝑣𝑗 = 𝜖𝑉𝑗 𝒕 /𝜖𝑡𝑗defines the marginal utility.
SLIDE 11 Dynamics of equilibrium
If a CP 𝑗’s profitability increases unilaterally
from 𝑤𝑗 to 𝑤𝑗
′, under Nash equilibrium, 𝑡𝑗 ′ ≥ 𝑡𝑗.
Dynamics of the Nash equilibrium:
𝜖s𝑗 𝜖𝑟 = 1 ⋯ 𝑗𝑔 𝑡𝑗 = 0 𝑗𝑔 𝑡𝑗 = 𝑟 𝑝𝑢ℎ𝑓𝑠𝑥𝑗𝑡𝑓 𝜖s𝑗 𝜖𝑞 = 0 𝑗𝑔 𝑡𝑗 = 0 𝑝𝑠 𝑡𝑗 = 𝑟 ⋯ 𝑝𝑢ℎ𝑓𝑠𝑥𝑗𝑡𝑓
SLIDE 12 Policy implications
Result: Under fixed price 𝑞, if marginal utility
matrix is off-diagonally monotone,
𝜖𝜚 𝜖𝑟 ≥ 0, 𝜖𝑆 𝜖𝑟 ≥ 0 𝑏𝑜𝑒 𝜖𝑡𝑗 𝜖𝑟 ≥ 0 ∀𝑗 ∈ 𝒪
- Deregulation incentivize CPs to subsidize,
increase system utilization and ISP revenue
Implications: deregulation is desirable for
improving investment incentives for ISPs
SLIDE 13 Policy under ISP’s optimal price
Consider a 3-stage game:
1.
Regulator chooses policy 𝑟
2.
ISP chooses optimal price 𝑞 𝑟
3.
CPs choose subsidies 𝒕 Policy effect:
𝑒𝑛𝑗 𝑒𝑟 = ⋯ , 𝑒𝜚 𝑒𝑟 = ⋯ , 𝑒𝜇𝑗 𝑒𝑟 = ⋯
CP 𝑗’s 𝜄𝑗 decreases with relaxed policy 𝑟 iff
𝜗𝑢𝑗
𝑛𝑗𝜗𝑟 𝑢𝑗/𝜗𝜚 𝜇𝑗 = 𝜗𝑟 𝑛𝑗/ 𝜗𝜚 𝜇𝑗 > −𝜗𝑟 𝜚 𝜗𝑢𝑗 𝑛𝑗 small: users are not price sensitive 𝜗𝜚 𝜇𝑗 large: traffic is sensitive to congestion 𝜗𝑟 𝑢𝑗 small: CP is less profitable
SLIDE 14
Relaxed policy induces higher 𝑆 and 𝑋 Price regulation might be needed
Revenue and social welfare
SLIDE 15 Conclusions
Study subsidization competition among CPs,
ISP uses the same per-unit charge Partial subsidy is allowed
Properties
the network is physically neutral it creates a feedback loop for CPs to compete increase access revenue and attract investment
Caveats
Utilization will increase, some CPs have lower rates ISP’s price might need to be regulated if the
market is not competitive enough
SLIDE 16 FCC Open Internet Order
Transparency
must disclose network management practices,
performance characteristics, and … No blocking
may not block lawful content, applications,
services, non-harmful devices … No unreasonable discrimination
may not unreasonably discriminate in
transmitting lawful network traffic …
SLIDE 17 How do we want to regulate?
It is about “no unreasonable discrimination” Existing solution
impose an absolute minimum requirement for
however, ISPs have different capacities …
Our proposal
restrict the maximum gap in service quality implication: if you make premium class better,
you need to make ordinary class better too.
SLIDE 18
References
Richard T. B. Ma. “Subsidization
Competition: Vitalizing the Neutral Internet.” ACM CoNEXT Conference 2014
Jing Tang and Richard T. B. Ma. “Regulating
Monopolistic ISPs Without Neutrality.” IEEE ICNP Conference, 2014.
