Your SlideShare is downloading. ×
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Internet pricing   time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Internet pricing time sensitive based data (visit http://trends-in-telecoms.blogspot.com for moe insights)

1,277

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
1,277
On Slideshare
0
From Embeds
0
Number of Embeds
7
Actions
Shares
0
Downloads
19
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® COMMUNICATIONS NETWORK ECONOMICS Incentivizing Time-Shifting of Data: A Survey of Time-Dependent Pricing for Internet Access Soumya Sen, Carlee Joe-Wong, Sangtae Ha, and Mung Chiang, Princeton University ABSTRACT congestion.1,2 Consequently, major U.S. ISPs like AT&T,3 Verizon,4 and Comcast5 have been aban- The tremendous growth in demand for broad- doning the traditional unlimited “flat rate” data band data is forcing ISPs to use pricing as a con- plans in favor of congestion-reducing mechanisms gestion management tool. This changing like throttling, data caps, and tiered usage-based landscape of Internet access pricing is evidenced (metered) pricing [3]. But such moves have been by the elimination of flat rate data plans in favor fraught with concerns for an open Internet, lead- of usage-based pricing by major wired and wire- ing to a polarizing debate on how ISPs should less operators in the US and Europe. But simple manage their network traffic to create a sustain- usage-based fees suffer from the problem of able Internet ecosystem. imposing costs on all users, irrespective of the This debate has primarily focused on the network congestion level at a given time. To issues of: effectively reduce network congestion, appropri- • Who should pay the price of congestion? ate incentives must be provided to users who are • What form should such pricing or network willing to time-shift their data demand from peak management policies take? 1M. Reardon, “AT&T to off-peak periods. These pricing incentives can The former question, which has serious net-neu- Wi-Fi Usage Skyrockets,” either be static (e.g., two-period daytime/night- trality implications [4, 5], relates to the issue of Cnet News, Oct. 24, time prices) or computed dynamically (e.g., day- whether content providers should pay a part of 2011. ahead pricing, real-time pricing). Data plans that their revenues to finance ISPs’ capital invest- offer such incentives to consumers fall under the ments in the underlying network infrastructure. 2E. Caggianni, “Growing category of time-dependent pricing (TDP). Many The latter question, which relates more directly Data Demands are Trou- ISPs across the world are currently exploring var- to the theme of this article, centers on the fair- ble for Verizon, LTE ious forms of TDP to manage their traffic growth. ness and appropriateness of current pricing poli- Capacity Nearing Limits,” This article first outlines the sources of today’s cies, such as application- and usage-based Talk Android, Mar. 6, challenges, and then discusses current trends pricing. App-based pricing, such as toll-free 2012. from regulatory and technological perspectives. (zero-rated) access to specific applications as Finally, we review representative pricing propos- practiced by Mobistar 6 in Belgium, is arguably 3 A. Lee, “AT&T to Begin als for incentivizing the time-shifting of data. discriminatory and much more contentious from Broadband Caps,” Huff- a net-neutrality standpoint [3]. However, usage- ington Post, May 2, 2011. INTRODUCTION based pricing (or data caps with metering) is widely viewed as an acceptable way to “match 4L. Segall, “Verizon Ends The demand for data is surging rapidly every price to cost,” even by the U.S. Federal Commu- Unlimited Data Plan,” year, with the Cisco Visual Networking Index for nications Commission (FCC) [6]. Yet the effec- CNN Money, July 6, 2012 projecting an 18-fold increase in global tiveness of usage-based pricing (UBP) has been 2011. mobile data traffic from 2011 to 2016. Mobile questioned by both academic and industry veter- video is predicted to be the fastest growing con- ans due to the fact that usage fees impose costs 5D. Goldman, “Comcast sumer mobile service, increasing from 271 million on users regardless of the actual network con- Scraps Broadband Cap, users in 2011 to 1.6 billion users in 2016 [1]. gestion in the network at any given time. For Moves to Usage-based Much of this demand growth is driven by the pop- instance, Vinton Cerf [7] wrote on Google’s Billing,” CNN Money, ularity of smart devices, bandwidth-hungry appli- Public Policy blog, “Network management also May 17, 2012. cations, cloud-based services, and media-rich web should be narrowly tailored, with bandwidth con- content [2]. This explosive growth in bandwidth straints aimed essentially at times of actual conges- 6 J. Cryderman, “Leverag- demand is now forcing Internet service providers tion. In the middle of the night, available capacity ing Policy 2.0 and Analyt- (ISPs) to invest in and expand their wired and may be entirely sufficient, and thus moderating ics,” Pipeline Magazine, wireless capacity by acquiring additional spec- users’ traffic may be unnecessary.” Andrew vol. 9, no. 2, July 2012, trum, deploying WiFi hotspots, and adopting new Odlyzko et al. [3] echoed a similar view in stating http://media.pipeline.pub- ______________ technologies such as fourth-generation (4G) Long that “Unless UBP contains a time-of-day billing spoke.com/files/issue/14/P ______________ Term Evolution (LTE) and femtocells. However, feature and some immediate feedback on conges- DF/PipelineJuly2012_A4. ______________ even these measures are proving insufficient for tion it is hard to imagine how it can be used as a __ pp. 3. pdf, meeting today’s challenges of network congestion management tool.” IEEE Communications Magazine • November 2012 0163-6804/12/$25.00 © 2012 IEEE 91C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 2. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® The widespread Growth in mobile data demand adoption of handheld devices, equipped with powerful processors, high-resolution cameras, and larger displays, has made it Mobile Cloud Data-hungry High-resolution Rapid growth in video services apps devices demand for mobile data convenient for Annual growth statistics users to stream Mobile video Total traffic (2011) 62% Streaming CAGR (2011 - 16) 34% high-quality videos Source: Cisco VNI 2011 - 2016 (http://www.tinyurl.com/VNI2012) and exchange large Cloud services Consumer traffic (2011) 83% Consumer traffic CAGR (2010 - 15) 67% volumes of data. Source: Cisco global cloud index 2010 - 15 (http://www.tinyurl.com/CiscoCloud2010) Data-hungry apps Facebook (2010) 267% Skype (2010) 87% Source: Allot global mobile trends report 2011 (http://money.cnn.com/2011/02/08/technology/smartphone_data_usage/index.htm) High-resolution devices iPad2 1024 x 768 pixels iPad LTE 2048 x 1536 pixels Source: Apple (http://www.apple.com/ipad/compare/) Figure 1. Some summary statistics on the annual growth of factors contributing to data explosion. Cerf and Odlyzko essentially argued that industry [2, 9, 10], and its potential as a viable UBP is not an appropriate pricing scheme for pricing policy for broadband data, this article congestion alleviation, as it does not address the reviews several TDP research works, with a par- problem of several users simultaneously access- ticular focus on incentive-based models for time- ing the network resources, which in turn leads to shifting of data. To contextualize the problem, large demand peaks and poor network perfor- we first provide a brief background of the key mance at those times. What ISPs can benefit factors that are contributing to network conges- from are smart data pricing7 (SDP) mechanisms, tion, followed by an analysis of the trends in which can create the right level of time-varying pricing policy changes that ISPs have adopted in incentives that induce consumers to shift their response and their regulatory implications. non-critical and bandwidth-heavy applications to periods of lower congestion. Time-dependent pricing (TDP) schemes focus on this aspect, and KEY DRIVERS OF hence can significantly help ISPs manage their networks. In fact, TDP for voice calls has been NETWORK CONGESTION widely practiced in the United States by AT&T Technological advances as well as changing con- and Verizon, with free calling on nights and sumer behavior are driving the explosion in data weekends but caps on minutes used during the demand. We discuss these factors next and pro- 7 Smart Data Pricing day. More dynamic versions of TDP have been vide some additional growth statistics in Fig. 1. (SDP) Forum, implemented by MTN Uganda and Uninor http://scenic.princeton.edu India, which update the prices for voice calls BANDWIDTH-HUNGRY DEVICES ______ SDP is a /SDP2012/. hourly, depending on the congestion conditions The widespread adoption of handheld devices, broader notion than time- at the call’s originating location [8]. Similar TDP equipped with powerful processors, high-resolu- shifting of data. It can be schemes for broadband data have been proposed tion cameras, and larger displays, has made it a combination of ideas, recently [2, 9] with a full system implementation convenient for users to stream high-quality including fair throttling, and trials [9, 10]. But it should be noted that videos and exchange large volumes of data. Data smart demand shaping, TDP does not necessarily need to have an explic- from laptops equipped with 3G dongles and net- application- or location- it pricing signal to the user; it can be implement- books with wireless high-speed data access con- dependent pricing, intelli- ed as an intelligent flat rate (IFR) scheme by tributes the most to wireless network congestion gent WiFi offloading, automating the client side to change consumer [2]. As for smartphones, Cisco projects that the quota-aware content dis- usage behavior in response to a feedback signal average monthly data usage will rise from 150 tribution, optimized con- from the network to stay within a given budget. MB in 2011 to 2.6 GB in 2016 [1]. New features tent caching and TDP leverages the trade-offs between con- like Siri on the iPhone 4S, which has doubled forwarding, and spon- sumers’ price sensitivity and delay tolerance Apple users’ data consumption, are driving this sored access. (time elasticity of demand) across different growth.8 applications to benefit both consumers and ISPs: 8J. Browning, “Apple’s consumers benefit from more flexibility of CLOUD SERVICE AND M2M APPLICATIONS Siri Doubles iPhone Data choice, while ISPs benefit from reduced costs of Cloud-based services that synchronize data Volumes,” Bloomberg capacity provisioning. Given the level of interest across multiple mobile devices, such as iCloud, News, Jan. 6, 2012. that TDP has received within both academia and Dropbox, and Amazon’s Cloud Drive, can be a 92 IEEE Communications Magazine • November 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 3. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® significant factor in traffic growth for ISPs and U-Verse and 150 Gbytes for DSL lines with $10 expensive monthly bills for consumers.9 Similar- overage for additional blocks of 50 Gbytes. SDP is in fact a ly, machine-to-machine (M2M) applications that For wireless networks, the transition to UBP generate data intermittently (e.g., sensors and has been faster. In 2010, AT&T introduced broad set of smart actuators, smart meters) or continuously (e.g., three tiered data plans of $15 fees for 200 mechanisms and video surveillance) often load the network with Mbytes, $25 for 2 Gbytes, and $45 for 4 Gbytes; ideas, which includes large signaling overhead [2]. Yet the intrinsic exceeding the caps resulted in $10/Gbyte over- time elasticity of such services also creates an age fees for the 2 and 4 Gbyte plans, and $15 for fair throttling, smart opportunity for shifting them to low-congestion each additional 200 Mbytes for the lowest tier demand shaping, times by applying appropriate price incentives. plan [3]. Within a year, AT&T also started throt- tling the heaviest 5 percent of its users, even intelligent WiFi CAPACITY-HUNGRY APPLICATIONS those with unlimited data plans [8]. As of July offloading, quota- The popularity of handheld devices has also led 2012, all major ISPs, including Comcast, AT&T, aware content distri- to rapid growth in the development of band- Verizon, and T-Mobile, have moved to some width-hungry applications for social networking, form of UBP; even Sprint, the only large carrier bution, sponsored file downloads, music and video streaming, per- with an unlimited data plan, cancelled it for lap- access, and so on. sonalized online magazines, and so on. For tops and notebooks.14 Next, we discuss some of example, Allot Communications reported that in the main criticisms and implications of replacing 2010, traffic from services like Skype grew about flat rate plans by UBP [3, 7]. 87 percent, and Facebook by 267 percent.10 Vir- gin Media Business reports that the average APPLICATION-BASED PRICING smartphone software uses 10.7 Mbytes/h, with Although highly contentious from a network the highest-usage app, Tap Zoo, consuming up neutrality standpoint, as discussed next, app- to 115 Mbytes/h.11 based pricing is emerging as a new trend in some These developments have led ISPs to view European and Asian markets [8]. App-based pricing as their ultimate congestion management pricing can come in several forms, such as zero- tool, paving the way for the adoption of SDP rating (toll-free) or sponsored content. Under measures, such as time-dependent and usage- these plans, either the ISP or content provider based pricing. pays for access to certain popular applications 9 D. Ngo, “iCloud: The (e.g., Mobistar, a Belgian ISP, has a zero-rate Hidden Cost for the plan for Facebook, Twitter, and Netlog). Anoth- Magic, and How to Avoid TRENDS IN PRICING FOR er form of app-based pricing is content and It,” Cnet News, Nov. 7, CONGESTION MANAGEMENT access bundling; for example, the Danish opera- 2011. tor TDC bundled its music streaming service, FLAT-RATE PRICING TDC Play, into its data plans [8]. In the United 10D. Goldman, “You are AOL’s historic switch 12 in 1996 from hourly States, cable provider Comcast recently came Using More Data Than rates to an unlimited flat rate data plan made under scrutiny for not counting its own Xfinity You Think,” CNN the latter the dominant model for access pricing. for Xbox video-on-demand service toward users’ Money, Feb. 8, 2011. However, as shown in Fig. 2, there has been a data caps. 15 We discuss the technological and drastic shift in ISPs’ pricing policies since 2008 regulatory perspectives of app-based bundling in 11K. C. Tofel, “Data in both wireline and wireless networks.13 Com- greater detail later. Hungry Mobile Apps Eat- cast’s data cap of 250 Gbytes was followed by ing into Bandwidth Use,” similar measures for AT&T’s U-Verse and digi- TIME-DEPENDENT PRICING Gigaom, May 19, 2011. tal subscriber line (DSL) services in 2011. Unlim- In the telecom industry, time-dependent pric- ited data plans for mobile services came to the ing has long been practiced by ISPs, often as 12M. Grebb, “Washington same end: between 2010 and 2012, AT&T, Veri- simple two-period (peak and off-peak) pricing State calls AOL on Flat- zon, and T-Mobile all eliminated their unlimited for voice calls. For example, both AT&T and Rate Plan,” Wired, Nov. plans in favor of tiered data plans. While AT&T Verizon, the two largest wireless operators in the 22, 1996. and Verizon embraced usage-based overages United States, offer free calling on nights and above data caps, T-Mobile adopted bandwidth weekends, while capping minutes used during 13 References to news throttling for users who exceeded their monthly the day. But today, operators in price-sensitive items discussed in this cap. regions such as India and Africa have begun to section can be found in offer more innovative dynamic TDP for voice Fig. 2. USAGE-BASED PRICING calls. MTN Uganda, for instance, has imple- Since 2008, the slow demise of flat-rate plans has mented a dynamic tariffing plan 16 in which the 14 J. O. Gilbert, “Sprint been widely covered in the media (Fig. 2). In price to make a voice call changes hourly Cancels Unlimited Data August 2008, Comcast implemented a 250 Gbyte depending on the congestion conditions at the for Hotspots, Tablets and cap on its cable network per household, threat- call’s originating location. While these TDP Netbooks,” Huffington ening repeat offenders (i.e., those who exceed innovations have been applied only to voice traf- Post, Oct. 21, 2011. the caps twice within six months) with termina- fic, there has been a system implementation and tion for one year [3, 8]. Time Warner conducted pilot trial of TDP for mobile data traffic [9, 10]. 15 D. Mitchell. “Is Com- UBP trials in Beaumont, Texas, which offered its cast Violating Net-Neu- customers a 768 kb/s line with 5 Gbyte cap for trality Rules?” CNN $29.95/mo or a 15 Mb/s line with 40 Gbyte cap REGULATORY AND RESEARCH Money, May 16, 2012. for $54.90/mo, both with $1/Gbyte overage fees. Although these met with poor reception, a simi- PERSPECTIVES ON PRICING TRENDS 16T. Standage, “The lar UBP trial by AT&T from November 2008 to As more innovative pricing practices are pro- Mother of Invention,” April 2010, conducted in Beaumont and Reno, posed by researchers and adopted by ISPs, they Special Report on Mobile Nevada, fared better [3]. By May 2011, AT&T should be analyzed with regard to their efficacy Marvels, The Economist, introduced a monthly data cap of 250 Gbytes on in meeting today’s challenges, as well as their Sept. 24, 2009. IEEE Communications Magazine • November 2012 93C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 4. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Wireless Introduction of Elimination of unlimited data plans usage-based penalties Verizon eliminates new unlimited smartphone plans No unlimited plans (July 2011)8 offered for iPad LTE (March 2012)10 AT&T starts throttling Verizon to eliminate unlimited iPhone users unlimited data plans (July 2011)9 (May 2012)11 AT&T starts charging a T-Mobile starts Verizon introduces $10/Gbyte overage fee data caps and shared data plans for smartphone data throttling penalty with unlimited plans (June 2010)6 (May 2011)7 voice and text (June 2012)12 2008 2009 2010 2011 2012 Time-Warner AT&T caps U-verse to 250 trials data Gbytes and DSL to 150 caps in Texas Gbytes with a $10 penalty for each additional 50 Gbytes (June 2008)1 per month (May 2011)4 AT&T starts Comcast moves Comcast caps toward tiered usage- data at 250 Gbytes throtting wireline based billing (August 2008)2 users (May 2012)5 (April 2011)3 Introduction of data caps Wireline Wireline 1. R. Paul, “40GB for $55 per month: Time Warner bandwidth caps arrive,” ArsTechnica, June 8, 2008. 2. Y. Adegoke, “Comcast to limit customer’s broadband usage”, Reuters, August 28, 2008. 3. B. Reed, “Bandwidth caps coming to AT&T wireline services,” Network World, March 14, 2011. 4. A.Lee, “AT&T to begin broadband caps,” Huffington Post, May 2, 2011. 5. D. Goldman, “Comcast scraps broadband cap, moves to usage-based billing,” CNN Money, May 17, 2012. Wireless 6. J. Cox, “AT&T shifts to usage-based wireless data plans,” Network World, June 2, 2010. 7. B. Reed, “T-Mobile’s data cap embrace leaves Sprint as lone ‘unlimited’ 4G carrier,” Network World, May 23, 2011. 8. L. Segall, “Verizon ends unlimited data plan,” CNN Money, July 6, 2011. 9. S. Lawson, “AT&T to throttle big users of unlimited data,” Computer World, July 29, 2011. 10. K.C. Tofel, “Apple’s iPad 4G LTE plan pricing detailed: Watch your speeds!” GigaOm, March 7, 2012. 11. J. Gilbert, “Verizon unlimited data plans for grandfathered customers to end soon,” Huffington Post, May 16, 2012. 12. R. Cheng, “Why Verizon’s shared data plan is a raw deal,” CNet News, June 12, 2012. Figure 2. Key events in the evolution of ISP pricing practices between 2008–2012. implications on a free and open Internet. This [11]. While UBP can be beneficial as a conges- section discusses academic research that seeks to tion control mechanism when growth in demand address some of the key questions and concerns exceeds the rate at which extra capacity can be related to pricing policies. An overview of the added, its effectiveness depends on its imple- key viewpoints on these different pricing plans is mentation [3]. Odlyzko et al. argue in [3, 12] that provided in Table 1. due to the statistical multiplexing principle of the Internet, bandwidth should not be viewed as IS USAGE-BASED PRICING USEFUL? a “consumable” good that is unfairly appropriat- ISPs like Comcast, AT&T, and Verizon have ed by bandwidth hogs. Moreover, they argue argued to the FCC that UBP is necessary for that the fiber deployment initiatives of AT&T investing in network infrastructure, so as to and Verizon were motivated more by competi- 17G. Fleishman, “The match their revenues to the cost incurred from tion with cable operators’ triple-play services State of 4G: It’s All About the explosive growth in data usage [6]. The opti- than by any real threat of congestion. In the case Congestion, Not Speed,” mal UBP mechanism that maximizes ISP rev- of wireless networks, where congestion at base Ars Technica, Mar. 29, enue in a resource-constrained network with stations is indeed an issue,17 the effectiveness of 2010. incomplete user information was presented in UBP is also under question [3, 7]. The key criti- 94 IEEE Communications Magazine • November 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 5. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Pricing Policy Arguments in Favor Arguments Against Adoption Trends Neutrality Perspectives • Modeling flat-rates as a form of bundling -shows the benefit of het- • Unsustainable because· of • Traditional model • Simpler, preferred by con- erogeneous user‘s preferences, “bandwidth hogs.” • Mostly discarded by US Flat-rate sumers despite disproportionate usage vol- • Light users subsidize heavy wired and wireless ISPs like (unlimited) • Users are typically more will- umes [12]. users. Comcast, AT&T, Verizon, T- ing to pay higher flat rate fees. • Flat pricing can cause loss of user • Encourages wastage. Mobile. net utility, particularly when the con- sumers have low price sensitivity [16]. • Data caps are less efficient than transmission caps [7]. • In 2008, Comcast intro- • UBP charges users regardless of duced a data cap of • Allows more efficient use of network congestion at any given 250GB/month per house- bandwidth. • Lacks the temporal dimen- time [3, 7]. Usage-based hold. • Can lower cost for consumers sion needed to solve ISPs’ peak • Network maintenance costs are not (UBP: data caps, • T-Mobile and AT&T throt- as they pay in proportion to congestion issue. directly relatable to traffic volume [3]. throttling, over- tle heaviest users. their consumption. • Can reduce consumer usage • ISPs’ revenue losses due to regu- ages) • Verizon and AT&T both • Helps ISPs to match their rev- and spending. lations against price discrimination eliminated flat-rate in lieu of enues to costs. can be mitigated by offering prices tiered data plans with nonlinear in data volume or data $10/GB overages. rate, with discounts on higher data-rate demand [16]. • Last-mile termination fees • Bundled subscriptions in imposed on third party content or • Offers consumers choices for • Discriminatory, non-neutral Europe by Tele-Danmark application providers will lead to personalized plans. • Favors big content and app Communications. paid prioritization over competing App-based • App bundling helps ISPs to providers. • Zero-rated access to Face- services, harming innovation and (zero-rating, gain consumers. • Deep Packet Inpsection for book, Twitter, Netlog by creating market inefficiencies [5]. toll-free apps) • Content-provider subsidies app classification has privacy Mobistar in Belgium. • Inability to price discriminate (toll-free) benefit consumers. implications. • Access and app bundling across flows causes higher revenue by Three in the UK. losses when users are less price sensitive [16]. • Exists as daytime/night- • Empowers users with the choice • Exploits time-elasticity of • Can penalize light, low- time discount. of scheduling elastic traffic to save Time-dependent demand by creating incentives income users who cannot shift • Already used in electricity, on monthly bills instead of flat-rate (TDP: static, for users to shift demand. their demand. transport markets. or volume caps [9, 10]. dynamic) • Can reduce peak demand and • Needs improved user inter- • Dynamic TDP for voice • Impact of TDP on application fill up valley periods. face for price notifications. calls is practiced in India, ecosystem needs further studies Africa. and trials [10]. Table 1. Summary of features and opinions on some current pricing policies. cism is that UBP cannot alleviate peak conges- arrangements in which app and content providers tion unless it contains a time-dependent pricing could pay ISPs to prioritize their packets over component to provide dynamic feedback on the competitors’. This would create market inefficien- network congestion level [10]. cies and hurt innovation, and can even incentivize ISPs to inflate network congestion problems in DOES APP-BASED PRICING order to extract higher revenues from content pro- THREATEN NET-NEUTRALITY? viders [5]. Similarly, agreements between ISPs and content providers on bundling applications and Reducing congestion requires investment in the access or zero-rating specific applications are network infrastructure, but questions remain on viewed as non-neutral, discriminatory practices [3]. whom to charge to subsidize this expense. Given that the consumers, even “bandwidth hogs,” CAN TIME-DEPENDENT PRICING impose a relatively small marginal operational HURT DEMAND? cost on the network, Odlyzko et al. [3] have argued that pricing mechanisms like UBP can One of the main arguments in favor of flat rate often be misused as a tool to increase return on pricing as opposed to UBP or TDP has been invested capital. On the other hand, former that consumers are willing to pay more for flat- AT&T CEO Ed Whitacre generated a new debate rate plans. Flat-rate plans encourage higher par- on net-neutrality by suggesting that in addition to ticipation (i.e., a positive network externality), connectivity fees, ISPs should be allowed to which benefits the Internet ecosystem [3]. Vin- charge content providers a part of ad revenues in ton Cerf suggests that time-dependent variants, exchange for the right to access end users.18 like UBP with off-peak discounts, can also “end Recent research by Musacchio et al. [4] has up creating the wrong incentives for consumers to shown that two-sided pricing (i.e., a “non-neutral” scale back their use of Internet applications over network) can sometimes increase social welfare, broadband networks” [7]. But in contrast to the 18 E. Whitacre, “At SBC, particularly when the ratio of certain model views expressed above, dynamic TDP for voice It’s All About “Scale and parameters characterizing advertising rates and calls by MTN Uganda actually increased the Scope,” edited by P. end-user price sensitivity is either high or low. In demand of their price-sensitive demographic.14 O’Connell, Business contrast to this view, Economides [5] has argued Broadband analytics firm Wireless Intelligence Week Mag., Nov. 6, that such ability would lead to paid prioritization reports that every one U.S. cent decrease in the 2005. IEEE Communications Magazine • November 2012 95C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 6. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Incentive Schemes Key Model Features Contributions Simulation/Systems Disadvantages • Higher prices and higher PLP can flatten out daily Simulations are used to Authors show that PLP arrival rate of user requests Peak load pricing demand and generates verify the effectiveness of can segment the user in pre-classified peak hours. (PLP) higher revenue than non- PLP in reducing peak uti- base, with low-budget • Elasticity of user requests (Parris et al. [13]) PLP, while also reducing lization in an integrated and low-elasticity users determines if the demand is call blocking probability. network. being denied service. deferred to off-peak hours. Studies the impact of off- The observations are Does not include a real Provides several simulation Off-peak discount peak window size, discounts, based on numerical world deployment or results on capacity savings (EI-Sayed et al. [2]) and usage shifted on the investigations for a trial results to validate or from peak load shifting. cost savings and revenues. range of scenarios. quantify the benefits. Optimized “day-ahead” TDP In trial results, the maxi- Requires careful design Dynamic day-ahead Conducts the first system for mobile data computed in mum peak to average ratio of user interfaces and pricing implementation and trial an ISP-user feedback control of usage volume decreases data usage notification (Ha et al. [9, 10]) of TDP for mobile data. loop. by 30%. system. • Dynamic adaptation of Provides analytical Requires automation for Real time pricing Introduces “Smart market” prices to congestion level. results, but does not end-users to bid on (MacKie-Mason and for real-time congestion • ISP admits packets based include a real trial of the packets during congest- Varian [14]) pricing. on user bids for auction. system. ed times. Game-theoretic • Introduces a parameterized Numerical examples Considers only a single Studies Price of Anarchy, analysis model of users choosing show cases when ISPs’ service provider scenario. i.e., social cost of ISP offer- (Jiang et al. [17]) their usage time based on optimal prices may be ing profit-maximizing congestion, price offers, and extremely socially subop- prices instead of maximiz- intrinsic preferences. timal. ing social welfare. (Shen and Basar • Uses non-cooperative Numerical results show Marketing of non-linear Shows that ISP profits can [15]) games to study user incen- ISP profits increase more pricing structure to con- be larger with non-linear tives from a mechanism than 38% with nonlinear sumers can be difficult. usage fees. design perspective. pricing over linear fees. Users receive a fixed number Shows that token pricing Numerical studies show Requires automation to Token bucket pric- of tokens that can be used to gives higher utility to users that token bucket pricing help users follow the ing obtain higher QoS during than flat rate for both sin- increases social welfare optimal token usage pol- (Lee et al. [18]) congested times. gle and two service classes. relative to flat rate. icy. ISPs need to measure Numerical studies show Lottery-based rewards in Shows that the equilibrium how much each user Raffle-based pricing robustness of social wel- proportion to users’ contri- prices maximize the total offloads, and user com- (Loiseau et al. [19]) fare to changes in user bution to peak reduction. social welfare. pensation depends on willingness to offload. lottery. Table 2. Comparison of features, contributions, and limitations in representative research works. effective price per minute of voice calls leads to ed periods. An overview of the features, contribu- a 6.9 and 13.5 times increase in minutes used tions, and limitations of the incentive models dis- per month for developed and developing coun- cussed below is provided in Table 2. tries, respectively. 19 With respect to mobile data, initial TDP trials (as discussed earlier) PEAK LOAD PRICING have found that TDP increases the overall Parris et al. [13] considered a time-dependent 19 http://www.wirelessin- demand, partly due to higher usage in discount- pricing model for peak load pricing (PLP) in ____________ telligence.com/analy- ____________ ed periods [9, 10]. which specified periods of time are pre-classi- sis/2011/09/how-pricing-d ______________ fied as peak and off-peak hours. In their simu- ynamics-affect-mobile- _____________ lation-based study, the authors consider a usage/ ___ INCENTIVE MODELS FOR higher arrival rate and a higher fee during the peak periods. Each user request has an associ- 20 Economic incentives TIME-SHIFTING OF DATA ated elasticity that determines whether a govern the actions of users Shifting demand from congested periods to off- request arriving during a peak period can be and ISPs, and forms a peak times requires pricing policies with the right deferred to a subsequent off-peak period. The core of the “economic incentives to induce such a response from users.20 authors show that PLP can spread out demand layer” of the network as Some of these policies (e.g., peak-load pricing, over time periods and generate higher revenues described by J. Walrand in off-peak discounts, day-ahead pricing, and real- than non-peak load pricing, but it can also seg- “Economic Models of time pricing) have an explicit time-dependent ment the user base, with low-budget and low- Communication Net- pricing feature; others, like game-theoretic (auc- elasticity users being denied service. This works,” Ch. 3, Perfor- tion) models and prioritization-based models potential adverse impact of PLP is similar to mance Modeling and (e.g., smart market, raffle-based, token bucket that of UBP (as observed by Odlyzko et al. in Engineering, Ed. Z. Liu, pricing), use more implicit time-varying incentives [3]) in that it hurts broadband adoption in low- Springer, 2008. to induce the time-shifting of data to less congest- income groups. 96 IEEE Communications Magazine • November 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 7. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Weighted average change in high price periods (%) 600 Auction-based mech- anisms help users to 500 50 explicitly specify their 0 400 -50 willingness to pay for -100 handling packets 300 -100 -50 0 50 from applications 200 with different time 100 elasticities. A limita- tion of this approach 0 is that it requires -100 intelligent automa- -100 0 100 200 300 400 500 600 700 800 Weighted average change in low price periods (%) tion on the client (a) (b) side to relieve the end users’ burden of Figure 3. TUBE Trial participants a) could view future prices (color-coded), along with price and usage his- tory. They responded by b) increasing their usage more in low-price, relative to high-price periods. decision making. OFF-PEAK PRICE DISCOUNTS GAME-THEORETIC PRICING Off-peak discounting from a predetermined Game theory offers a natural, but somewhat styl- baseline price is a logical dual of peak-load pric- ized, model for time-dependent pricing, as it ing. El-Sayed et al. [2] numerically studied time- allows one to consider consumers’ utility in the dependent price discounts for mobile data. Their amount of usage shifted in response to offered model includes three parameters to quantify the prices. Shen and Basar [15] use non-cooperative impact of TDP on both network cost savings and games to study user incentives from the perspec- revenue loss: tive of mechanism design. Although their work • The duration of the off-peak window in does not consider time-dependent pricing, they which off-peak incentives are offered show that while most service providers use linear • A discount percentage per megabyte (i.e., usage-based) charging, nonlinear charging • The percentage of busy hour load that the can yield large increases in ISP profit. Their users shift due to off-peak incentives results are relevant in the context of observa- The basic idea of offering TDP off-peak dis- tions made by Hande et al. [16] that nonlinear counts is similar to later works on dynamic TDP pricing can also reduce ISPs’ revenue losses, [9, 10], though these latter works also include an even in regulatory environments that prohibit analytical model of a control-feedback loop that explicit price discrimination across consumers. adjusts the offered discounts and presents a real- Jiang et al. [17] study hourly time-dependent world trial. pricing offered by a selfish ISP, comparing the profit-maximizing time-dependent prices to the REAL-TIME PRICING socially optimal ones. To compute these prices, The time-shifting of demand can also be real- the authors assume that users choose the timing ized by providing end users with an explicit of their Internet usage based on the congestion pricing signal about the real-time network con- condition, price offered, and intrinsic preference dition. MacKie-Mason and Varian [14] present for that time slot. They then focus on the price the idea of a “Smart Market,” which is a closed of anarchy, defined in terms of the social welfare control-feedback loop in which the network under profit-maximizing time-dependent prices adapts the prices to congestion levels. In this when compared to the maximum social welfare. approach, each user places a “bid” on each If several low-utility users are present in the net- packet that reflects her willingness to pay to work and the ISP cannot offer different prices to send the packet onto the network at a given these users, the price of anarchy may be arbitrar- time. The gateway admits packets in descending ily large (i.e., the ISP-optimal prices may be order of their bids as long as the network per- extremely socially suboptimal). This result sug- formance remains above a desired threshold. gests that some regulatory price restrictions can Users are charged according to the minimum improve social welfare. bid on a packet admitted into the network at the time, and thus pay only for the congestion TOKEN BUCKET PRICING cost at the market-clearing price. Auction- Token bucket pricing [18] divides the day into based mechanisms help users to explicitly speci- peak and off-peak hours. Users can accumulate fy their willingness to pay for handling packets daily tokens, which may be exchanged for service from applications with different time elastici- during peak hours; without sufficient tokens, ties. A limitation of this approach is that it users must wait until off-peak hours to consume requires intelligent automation on the client data. Viewing the tokens as currency, one sees side to relieve the end users’ burden of decision an analogy with time-dependent pricing [2, 9, making. 14]; the principal difference is that users’ bud- IEEE Communications Magazine • November 2012 97C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 8. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® gets with token pricing are determined by the and off-peak discounts [2] is its dynamic adjust- TUBE is the first fixed rate at which tokens are received, instead ment of future prices in response to changes in of the variable amounts of money they are will- usage behavior; TUBE thus performs better in functional TDP ing to spend on peak-hour usage. The presence reducing peak congestion [10] than does static system for mobile of these token-based incentives is shown to TDP. Additionally, to the best of our knowledge, data that has increase social welfare relative to flat-rate unlim- TUBE is the first functional TDP system for ited data plans. However, token pricing may mobile data that has undergone a real customer undergone a real require client-side automation for users to follow trial 22 [9, 10]. The TUBE project has also customer trial. “optimal” policies for using their tokens. More- released a free personalized bandwidth manage- over, token-based pricing may require several ment app, called DataWiz, for iOS and Android different plans, corresponding to different rates devices for the general public (http://scenic. of token accumulation. For instance, a business princeton.edu/datawiz/). user may be willing to pay a higher flat fee in exchange for more tokens that can be used dur- ing congested hours of the day. FEASIBILITY STUDIES As discussed in the previous sections, incentiviz- RAFFLE-BASED PRICING ing time-shifting of data has been practiced in Raffle-based pricing also divides the day into the form of static peak and off-peak discounts two time periods: peak and off-peak. Users are for voice calls. But more dynamic forms of TDP then offered probabilistic incentives to shift their are emerging in growing economies; MTN Ugan- demand to off-peak periods, in the form of a raf- da’s dynamic tariffing plan for voice calls result- fle or lottery for a given monetary reward. The ed in a volume increase of 20–30 percent, probability of winning the lottery is proportional indicating that price incentives for voice calls can to the user’s contribution to the reduction in strongly influence user behavior. peak demand. This “all-or-nothing” lottery may The potential benefits of TDP can be higher in instead be replaced by one in which the ISP the case of mobile data, since many applications divides the total reward by the total amount of today (e.g., online backups in the cloud, app and traffic shifted, paying this amount to each user movie downloads, peer-to-peer) have an intrinsic with a probability equal to the percentage of time elasticity that can be effectively exploited usage shifted to the off-peak period. In [19], [10].23 TUBE successfully induced consumers to Loiseau et al. derive expressions for the Nash shift their data demand to lower-priced, less con- equilibrium of this user-ISP interaction and gested periods [9, 10]. This nine-month trial con- show that in some cases the social welfare is sisted of 50 participants with iPads or iPhones and more robust to price variations than a compara- AT&T 3G subscriptions. Trial participants were ble time-of-day pricing plan with two periods. charged for mobile 3G data usage according to Yet the uncertainty inherent in raffle-based pric- TUBE’s day-ahead pricing algorithms. As shown in ing may reduce its effectiveness; users may not Fig. 3a, participants could view their price and shift their usage to off-peak periods without usage history, as well as future prices, on their iOS guaranteed rewards. This uncertainty is exacer- devices. In one phase of the trial, the hourly prices 21 CNET Energy Pricing. bated by the reward’s dependence on other alternated between high and low, to test users’ http://www.cntenergy.org/p users’ behavior; hence, a field trial of such a pol- price sensitivity. Figure 3b shows the resulting ricing/ ___ icy would be necessary to understand its efficacy. change in usage, measured relative to usage with static UBP of $10/Gbyte. The reference line shows 22 The Berkeley INDEX DYNAMIC DAY-AHEAD PRICING an equal percentage change in usage for low- and project of the 1990s (P. P. Ha et al. [10] introduce Time-Dependent Usage- high-price periods, with each data point corre- Varaiya, R. J. Edell, and Based Broadband Price Engineering (TUBE; sponding to the weighted average of one user’s H. Chand, INDEX Pro- http://www.datami.com), a fully integrated TDP percentage changes in usage in all high-price and ject Proposal, 1996) system that uses “day-ahead pricing,” a scheme all low-price periods. Since most data points lie attempted a similar initia- often used in electricity markets.21 In this dynam- below the reference line, most users increased tive for wired network ic TDP policy, the prices vary by the hour, but their usage in low-price relative to high-price peri- pricing. users are informed of the prices one day in ods, indicating some price sensitivity. When opti- advance. This advance notice alleviates the user mized time-dependent prices were offered, this 23 Dynamic day-ahead uncertainty inherent in real-time pricing, allow- price sensitivity helped bring down the peak pricing has also been suc- ing users to plan their bandwidth consumption demand, measured in terms of the peak-to average cessfully practiced for over the next day if desired. TUBE models this ratio (PAR) over each day. The maximum PAR some years in the electrici- user behavior as shifting mobile data usage from decreased by 30 percent, while the overall usage ty industry, e.g., Ontario one period to another in response to future increased by 130 percent due to higher demand in IESO prices. Users have a certain probability of shift- the valley periods with large discounts [10]. (http://www.ieso.ca/imow ing their traffic from each period to each other eb/marketData/market- period, as a function of the time difference _____________ _____ Viewing differ- Data.asp). between the periods and the price in the later CONCLUSIONS ent mobile applications as period. Given functions describing these proba- The growing fear of a data tsunami, particularly analogous to appliances bilities, the ISP can calculate the traffic pattern in wireless networks, has led ISPs to use pricing with different delay toler- over the day as a function of the prices offered as a congestion management tool. However, ances, one sees a clear and use this calculation to optimize its profit. understanding the potential benefits and risks analogy between the These prices are offered to users on their indi- associated with different pricing schemes is cru- demand-response models vidual devices; the ISP then records the resulting cial in both choosing effective data plans and of time-dependent pricing usage and revises its estimates of users’ respons- guiding regulatory decisions in maintaining an for mobile data and elec- es to offered prices. open Internet. For more than two decades, tricity market. TUBE’s principal advantage over PLP [13] research in network economics has shown that 98 IEEE Communications Magazine • November 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  • 9. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® providing the right incentives for consumers to [14] J. MacKie-Maso and H. Varian, Pricing the Internet, Public Access to the Internet, B. Kahin and J. Keller, time-shift their demand, as commonly done with Eds., Prentice-Hall, 1995. The growing fear of time-dependent pricing, is the key to resolving [15] H. Shen and T. Basar, “Optimal Nonlinear Pricing for a the issue of network congestion. In this work, we Monopolistic Network Service Provider with Complete a data tsunami, par- provide an overview of various incentive mecha- and Incomplete Information,” IEEE JSAC, vol. 25, no. 6, ticularly in wireless 2007, pp. 1216–23. nisms for the time-shifting of demand, including [16] P. Hande et al., “Pricing under Constraints in Access networks, has led some results from very recent works and con- Networks: Revenue Maximization and Congestion Man- sumer trials, with the aim of helping researchers agement,” Proc. IEEE INFOCOM, San Diego, CA, 2010. ISPs to use pricing as and practitioners become aware of the various [17] L. Jiang, S. Parekh, and J. Walrand, “Time-Dependent Network Pricing and Bandwidth Trading,” Proc. IEEE a congestion man- challenges and opportunities arising from today’s NOMS Wksp., Salvador, Brazil, 2008. network congestion problems. The evolution of agement tool. How- [18] D. Lee et al., “A Token Pricing Scheme for Internet new technologies like 4G/LTE, smart grids, and Services,” Proc. 7th ICQT, Paris, France, 2011. ever, understanding cloud computing will greatly depend on the pric- [19] P. Loiseau et al., “Incentive Schemes for Internet Con- gestion Management: Raffle versus Time-of-Day Pric- the potential benefits ing policy choices that we make today. ing,” 2011 Annual Allerton Conf., Urbana-Champaign, IL, 2011. and risks associated ACKNOWLEDGEMENTS with different pricing We would like to specially thank Andrew BIOGRAPHIES Odlyzko, Roch Guerin, Nicholas Economides, schemes is crucial. SOUMYA SEN [S’08, M’11] received a B.E. (Hons.) in electron- Rob Calderbank, Krishan Sabnani, Shyam ics and instrumentation engineering from BITS-Pilani, India, Parekh, Sundeep Rangan, Victor Glass, in 2005, and both M.S. and Ph.D. degrees in electrical and Prashanth Hande, Raj Savoor, Steve Sposato, systems engineering from the University of Pennsylvania in Patrick Loiseau, and all the participants of the 2008 and 2011, respectively. He is currently a postdoctoral research associate at Princeton University. He has been 2012 Smart Data Pricing (SDP) workshop in actively involved in forging various industrial and academic Princeton, for sharing their valuable comments partnerships, and serves as a reviewer for several ACM and and insights. We also acknowledge the support IEEE journals and conferences. He is also a co-organizer of of our industrial collaborators, in particular, the Smart Data Pricing (SDP) Forum 2012 and GLOBECOM 2012 SDP Industry Forum, which promotes industry-aca- Reliance Communications, MTA Alaska, AT&T, demic interaction on pricing policy research. His research Qualcomm, Comcast, Mediacom, and the interests are in Internet economics, communication sys- National Exchange Carrier Association. tems, social networks, and network security. He has pub- lished several research articles in these areas and won the Best Paper Award at IEEE INFOCOM 2012. REFERENCES [1] Cisco Visual Networking Index, White Paper, Feb. 2012, CARLEE JOE-WONG [S’11] is a graduate student at Princeton available online at http://www.cisco.com/ en/US/solu- University’s Program in Applied and Computational Mathe- tions/collateral/ns341/ns525/ns537/ns705/ns827/white_p ______________________________ matics. She received her A.B. (magna cum laude) in mathe- aper_c11-520862.pdf, 2012. ____________ matics from Princeton University in 2011. Her research [2] El-Sayed et al., “Mobile Data Explosion: Monetizing the interests include network economics and optimal control Opportunity through Dynamic Policies and QoS Pipes,” theory, and her work on the fairness of multi-resource allo- Bell Labs Tech. J., vol. 16, no. 2, 2011, pp. 79–100. cations received the Best Paper Award at IEEE INFOCOM [3] A. Odlyzko et al., “Know Your Limits: Considering the 2012. In 2011, she received the NSF Graduate Research Fel- Role of Data Caps and Usage Based Billing in Internet lowship and the National Defense Science and Engineering Access Service,” Public Knowledge White Paper, Graduate Fellowship. http://publicknowledge.org/files/UBP%20paper%20FINA L.pdf, 2012. ___ S ANGTAE H A [S’07, M’09, SM’12] is an associate research [4] J. Musacchio, G. Schwartz, and J. Walrand, “A Two- scholar in the Department of Electrical Engineering at Sided Market Analysis of Provider Investment Incentives Princeton University, leading the establishment of the With an Application to the Net-Neutrality Issue,” Rev. Princeton EDGE Laboratory as its associate director. He Network Economics, vol. 8, no. 1, Berkeley Electronic received his Ph.D. in computer science from North Carolina Press, 2009. State University in 2009 and has been an active contributor [5] N. Economides, “ Why Imposing New Tolls on Third- to the Linux kernel. During his Ph.D. years, he participated Party Content and Applications Threatens Innovation in inventing CUBIC, a new TCP congestion control algo- and Will Not Improve Broadband Providers’ Invest- rithm. Since 2006, CUBIC has been the default TCP algo- ment,” NET Institute Working Paper No. 10-01; NYU rithm for Linux and is currently being used by more than Law and Economics Research Paper No. 10–32, 2010. 40 percent of Internet servers around the world and by [6] S. Higginbotham, “FCC Opens the Door for several tens of million Linux users for daily Internet com- Metered Broadband,” Dec. 1, 2010, http://gigaom.com/ munication. His research interests include pricing, greening, 2010/12/01/fcc-opens-the-door-for-metered-web- ______________________________ cloud storage, congestion control, peer-to-peer network- access/, 2010. ____ ing, and wireless networks. [7] V. Cerf, “What’s A Reasonable Approach for Managing Broadband Networks?,” Google Public Policy Blog, M UNG C HIANG [S’00, M’03, SM’08, F’12] is a professor of August 4, 2008, http://googlepublicpolicy.blogspot. electrical engineering at Princeton University, and an affili- com/2008/08/whats-reasonable-approach-for-manag- ______________________________ ated faculty member in applied and computational mathe- ing.html, 2008. _____ matics, and computer science. He received his B.S. (Hons.), [8] S. Sen et al., Pricing Data: Past Proposals, Current Plans, M.S., and Ph.D. degrees from Stanford University in 1999, and Future Trends, http://arxiv.org/pdf/1201.4197.pdf, 2000, and 2003, respectively, and was an assistant profes- 2012. sor 2003–2008 and associate professor 2008–2011 at [9] S. Ha et al., “Demo: Enabling Mobile Time-Dependent Princeton University. His research on networking has Pricing,” Proc. ACM Mobisys, Lake District, U.K., 2012. received a few awards, including the 2012 IEEE Kiyo [10] S. Ha et al., “TUBE: Time Dependent Pricing for Mobile Tomiyasu Award, a 2008 U.S. Presidential Early Career Data,” Proc. ACM SIGCOMM, Helsinki, Finland, 2012. Award for Scientists and Engineers, several young investi- [11] S. Li, J. Huang, and S.-Y. R. Li, “Revenue Maximization gator awards, and several paper awards. His inventions for Communication Networks with Usage-Based Pric- resulted in seven U.S. patents and a few technology trans- ing,” Proc. Globecom, Honululu, HI, 2009. fers to commercial adoption, and he received a Technology [12] P. Nabipay, A. Odlyzko, and Z.-L. Zhang, “Flat Versus Review TR35 Award in 2007 and founded the Princeton Metered Rates, Bundling,” and “Bandwidth Hogs,” EDGE Laboratory in 2009. He serves as an IEEE Communi- Proc. NetEcon, San Jose, CA, 2011. cations Society Distinguished Lecturer in 2012–2013, and [13] C. Parris, S. Keshav, and D. Ferrari, A Framework for wrote an undergraduate textbook, Networked Life: 20 the Study of Pricing in Integrated Networks,Tech. Rep., Questions and Answers, for a new undergraduate course, Tenet Group, ISCI, UC Berkeley, 1992. Networks: Friends, Money, and Bytes. IEEE Communications Magazine • November 2012 99C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®

×