There has been no shortage of headline news on Internet access pricing in 2010 and 2011. From AT&T and Verizon shifting to usage-based pricing for wireless access to FCC's revealing the National Broadband Plan, and from Comcast-Netflix/Level3 battle on two-sided pricing to FCC's 2010 December statement on pricing innovations, the Internet is witnessing the start of a transformative period in the interplay between access pricing and the technology of networking.

The technical core of this project is organized around four fundamental questions: (1) How much to charge? The debate between flat-rate and usage-based pricing renews with fresh perspectives. (2) How to charge? Should the price charged depend on the time of bandwidth consumption or congestion condition of the network? (3) Whom to charge? When will content/application producers find incentives to pay for higher data rate or heavier bandwidth consumption by consumers? (4) What to charge? What kind of new service classes can be invented, especially in heterogeneous wireless networks with multiple platforms co-existing?

The unique features of this project include extensive participation from diverse sectors in the networking industry, implementation of prototypes and trials, and access to and dissemination of data. It combines the collection of fresh, large-volume of empirical data with rigorous design and prototype implementation. The project goes all the way from data analysis through optimization algorithms, to proof-of-concept demos and trials, eventually to practical impact on public policy and ISP business decisions, thus closing the loop in the study of network pricing.

Intellectual Merits: While we sharpen and apply a variety of tools from optimization theory, microeconomics, game theory, and statistics, the challenges arising out of these social issues also demand the development of new methodologies, such as supply chain contracts for multi-platform, two-sided pricing with conflicting interests of content providers and Internet service providers (ISPs). Furthermore, the interactions between technology evolution and economic policies are mutual: new enabling technologies such as femtocell raises new questions on the interaction between engineering artifacts and pricing structures.

Broader Impacts: The proposal is driven by timely and important questions faced by policy-makers, networking industry, and broadband consumers: (1) Who will pay for the estimated cost of $350B in the next decade to enable universal coverage of broadband services in this country? (2) Can pricing mechanisms be leveraged by the ISP as a practical approach to network management and new service class be created that is net-neutrality compatible? (3) How to regulate the nonstop surge of bandwidth demand to create win-win for both the ISPs and consumers? Furthermore, the project presents unique opportunities for undergraduate curriculum development, extensive industry participation and impact, and unconventional community outreach, both within US and across the world.

Project Report

November 2014 This project examined different ways for Internet service providers (ISPs), possibly those outside of US, to charge their users for consuming data. We examined five different pricing plans: time-dependent pricing, WiFi offloading, bundled pricing of different network types (e.g., cellular and WiFi), sponsored content pricing, and traded data plans. Our work is motivated by the growing volume of mobile data traffic, which is increasingly causing noticeable congestion on cellular networks, especially at times of peak usage. ISPs have started offering different Smart Data Pricing (SDP) plans in the recent years in different countries. Without presuming any particular policy stance, this project focuses on the fundamental research of network economics and technologies. For example, Time-dependent pricing (TDP) charges users different prices at different times of the day, for instance $10/GB for nighttime and $15/GB for daytime usage. By offering lower prices at less congested times, ISPs encourage users to shift their usage to cheaper, less congested times of the day, thus reducing peak congestion. We investigated a form of TDP called day-ahead pricing, in which the prices are set 24 hours in advance. Users can thus view the prices for the next day and plan their usage in advance. We developed algorithms to optimally compute day-ahead prices and designed and implemented a system for offering these prices to users. We then conducted a pilot trial of TDP for mobile data: we acted as a resale ISP to users, charging them for usage according to the time-dependent prices. We found that users do shift their usage to lower-price periods, which enables a 30% decrease in peak usage relative to average usage. Moreover, users may additionally increase their usage at lower-price times, due to a "sales-day" effect from observing the lower prices (perceived as "discounts" off the prices charged at other times). ISPs can also reduce peak-time cellular usage by introducing WiFi offloading algorithms. Since WiFi is generally free and often offers higher speeds than cellular networks, users have an incentive to wait for WiFi access. Some apps, however, cannot wait, e.g., GPS directions while driving. We develop algorithms to automatically encourage some apps to wait for WiFi depending on users’ preferences for delay, higher speeds, and usage costs. Experiments on cellular and WiFi usage data from 37 users show that users on average save 30% with our algorithms. ISPs can encourage offloading by offering users bundled data plans that include access to their WiFi network as well as cellular networks. We consider the optimal prices for two different pricing plans: access to only the cellular network and access to both cellular and WiFi networks (a bundled plan). We show that simple strategies for encouraging WiFi usage can lead to unintended consequences. For instance, expanding the WiFi coverage area can lead to more congestion on WiFi networks, leading to some users dropping the bundled data plan. Sponsored content pricing aims to increase ISP revenue by charging content providers (CPs) as well as users for data usage: CPs can choose to subsidize users’ data delivery costs. We consider the effect of sponsored data on different types of users and CPs. We find that sponsored data disproportionately benefits low-income users, indicating that it can help increase access to mobile data by boosting usage for users who otherwise could not afford to pay for mobile data usage. Traded data plans assume that users have caps on their monthly data usage. We consider an ISP that allows users to sell leftover data caps to other users who might have exceeded their monthly caps. Buyers and sellers submit "bids" with the amount of data they wish to buy or sell and the price that they wish to pay or receive; the ISP then matches buyers to sellers so as to maximize its revenue. We explored conditions under which both ISPs and users can benefit from the choice of such a market. The project has led to many outreach activities. A book called Smart Data Pricing was published in September 2014, and a book chapter on the same topic was published online in 2013 in the open-source SIGCOMM E-Book on Advanced Technologies in Networking. Multiple survey articles have been published on SDP. A startup on SDP solutions was spun-out from the university. We also hosted three workshops on SDP featuring industry and academic speakers.

Agency
National Science Foundation (NSF)
Institute
Division of Computer and Network Systems (CNS)
Type
Standard Grant (Standard)
Application #
1117126
Program Officer
Darleen L. Fisher
Project Start
Project End
Budget Start
2011-09-01
Budget End
2014-08-31
Support Year
Fiscal Year
2011
Total Cost
$300,000
Indirect Cost
Name
Princeton University
Department
Type
DUNS #
City
Princeton
State
NJ
Country
United States
Zip Code
08540