Dennis Weller, Senior Advisor, Navigant Economics "IP Traffic Exchange; Market Developments and Policy Challenges"
IP Traffic Exchange Market Developments and Policy Challenges BEREC/OECD Workshop Brussels, 2 November 2011 Dennis Weller Navigant Economics email@example.comPage 1
Growth of the Internet» Two billion users» Eightfold traffic growth in five years › Still dramatic, but less than in previous periods» Global extension of the Internet › High rates of growth in eastern Europe and emerging economiesPage 5
Growth in Bandwidth Production by Country and Region » Five-year annualized growth ranges from less than 20% to › 1,470% (Russia) › Over 400% (South Africa, Uganda, Egypt) Annualized percentage growth in domestic Internet bandwidth production, grouped by region, 2005–2010
Geographic Expansion of the Internet » Investment in IXPs (now about 350) » Domestic production capacity promotes economic self-sufficiency » Local IXPs reduce need for “tromboning” of traffic » IXPs provide centers for development of Internet assets within region » 53% of countries, including many LDCs, do not yet have an IXP
Well-Understood and Accepted Norms for Peering » A survey of 142,210 peering agreements was conducted for the OECD report. » 86% of Internet carriers represented, in 96 countries, including all of OECD » 99.51% of agreements required no written contract » 99.73% had symmetric terms » High prevalence of multilateral agreements » US, Canada, Japan, Singapore were most often selected governing jurisdictions
Performance of the Market» Low prices › Wholesale transit at USD 2-3 per megabit per second per month › Rate varies by distance from IXPs and by volume purchased › Equivalent to USD 0.0000004 per voice minute › Roughly constant since 2006: supply/demand ratio worsening at the same time that technological progress drives some input costs down» Universal connectivity with relatively sparse agreements › Less than 1% of a full mesh › Average AS-path length is currently 4.3 › Maximum AS-path length, corrected for prepending, is currently 12.
Structural Change in Internet Exchange Model » Reduced role for traditional incumbent networks › High degree of “donut peering” among Internet-centric carriers › Disintermediation of transit › Traditional incumbents marginalizing themselves through lack of interconnection » 10,000 Traditional » 100,000Incumbents » 0 100Prefixes Served Degree of Interconnection
Structural Developments in Services» Voice declining in importance › Already a tiny fraction of traffic exchanged › Voice usage declining in all age groups ‒ US Teenagers (ages 13-17) make or receive an average of 3,339 texts per month (Nielsen)» P2P still significant, but share no longer increasing › P2P 40% of traffic in 2010, 16% by 2015 (Cisco) ‒ But still the greatest source of demand for upload capacity» Growth from: › Streaming, direct downloads of video ‒ Fastest growing category 2007-2009 (Atlas) › Applications moving from local storage to online › Supported by growth of CDNs
Structural Developments in IP Exchange Market» Growth of Content Delivery Networks (CDNs) › Stand-alone providers ‒ Akamai’s revenues have quadrupled in last five years ‒ In 2009, the top five stand-alone CDNs carried almost 10% of Internet traffic › Self-supply by online content providers like Google, which now carries 6% of Internet traffic › Investment in direct transport, caching › Improved quality; alternative to transit» Boundaries blurring between content provider, backbone, local access, CDN, as all are providing these functions» Part of overall evolution toward more direct routing and improved quality
Challenge: Deviation from Moore’s LawExponential growth in Interface speeds made possible by research in optoelectronicphysics.After bubble burst in 2000, investment in research dried up -- with a lag, this led todeviation from previous trend. 115% annual exponential 6 gbps / year linear Interface speeds available at IXPs, by year
When Worlds Collide -- New and Old Models IP Model Traditional ModelNot Regulated RegulatedGrowin g D ec l i n ing nce!Better Worse rm a Pe r f o Perfo rman ce
Policy Challenges» Convergence -- When Worlds Collide › IP and TDM models create different terms of trade, incentives › “Bright line” between the two frameworks › Important to avoid applying old frameworks to IP, in both domestic and international policy» The question of voice» Reform of TDM traffic exchange» De-privatization» Pricing models to support new investment in access networks» Network neutrality and the IP model of traffic exchange
Evolution of the Market » All players in the value chain reinventing business models, terms of trade » New variations on the model of Internet traffic exchange › Multilateral peering › Regional peering › Single-hop or backplane access › CDN access or paid peering » Traffic balance -- what role should it play? » Transit, peering, and CDN as substitutes
Pricing Options for Local Access Networks » Growth in traffic creates demand for investment by local access network › Shift in delivery mode › Depends on how traffic is delivered ‒ How deep in access network ‒ Caching › Need to distinguish incremental demand associated with online content delivery from more general upgrade investment » Four options suggested by AT Kearney › Non-linear end-user prices ‒ Linkage with wholesale agreements? › Mandatory terminating charge › Coordination for QOS › Provision of CDN functions
Network Neutrality and the Internet Market for Traffic Exchange» Many of the questions raised in the network neutrality debate are now being answered in the market for IP agreements › How should Internet content be delivered to broadband users? › What resources should be deployed to improve quality, where, by whom, and on what terms?» Recent disputes in the headlines: › Cogent/Orange › Level 3/Comcast» Current state of play: › Results are mixed, but fall within a reasonable range › Google has peering agreements with 70% of Internet providers › Where paid peering has been agreed, charges are on the lower end of range for transit › Negotiation of considerations in money and in kind › Simple peering model a low-cost alternative