4. Staying Interconnected
IP interconnection agreements are the lubricating oil of internet infrastructure. These are the commercial arrangements that dictate how traffic is passed among the thousands of networks that make up the internet. As the sheer volume of digital traffic has soared in recent years, driven in large part by video, these agreements have come under increasing pressure for renegotiation and oversight by public authorities in a number of countries. Other sources of expanding data will only add to this pressure. Everyone has an interest in their continued smooth operation, because everyone benefits from the continuing efficient flow of data.
Net neutrality and IP interconnection agreements are related but distinct issues. Both concern how traffic is managed across networks, but at different points. The debate over net neutrality affects the so-called “last mile” – the connection between the internet and the end-user. (See the sidebar, “IP Interconnection and Net Neutrality”.)
The Short History of IP Interconnection Agreements
IP Interconnection refers to the commercial agreements among network providers that exchange the traffic transmitted across the internet. There are three principal types of agreements:
- Settlement-free peering, in which two network providers agree to accept from each other, free of charge, traffic that terminates on each party’s network. The basis for this exchange is that each operator receives mutual value, so each pays to maintain the infrastructure necessary to support traffic levels on its own network. With the vast amount of traffic that is exchanged every day, settlement-free peering simplifies these transactions and provides for efficient traffic routing.
- Paid-peering agreements, under which one network (or, more recently, a network operated by a content provider) pays another to terminate traffic on its network. These are much less common than settlement-free peering agreements.
- Transit agreements, under which a party pays a network provider to accept traffic that is destined to or from another network anywhere on the internet. (See Figure 6.)
Figure 6: Illustration of Common IP Interconnection Agreements
Note: Simple depiction of typical IP Interconnection agreements; many other variations can and do exist
Source: BCG analysis
All three types of IP interconnection agreements are wholesale arrangements, usually struck between network operators; they do not involve consumers. In the past, they did not directly involve content providers either, but the lines are blurring. In February 2014, Netflix struck an IP interconnection agreement with US cable operator Comcast, under which Netflix will load its streaming video content directly onto Comcast’s network.17 In addition, some large cloud providers have effectively become private network providers because they operate their own networks and negotiate IP interconnection agreements with other networks for their own traffic to and from their data centres.
IP interconnection agreements have evolved on a global basis since their inception. To date, most of the traffic travelling among network providers has been exchanged without charge in settlement-free peering agreements.18 Until recently, most of the peering traffic ratios have remained within relatively narrow (and similar) bands that reflected reciprocal traffic levels, 1:5, for example, or 5:1. In recent years, however, disputes have arisen among CSPs as the volume of content carried has soared, and some peering ratios have become lopsided, such as 1:10 or 1:20, because of the rapid growth in content providers’ services and the associated traffic (predominantly video, but likely to include more games and other media in the future). In the US today, Netflix and YouTube together account for more than half of all downstream fixed broadband traffic.19 Some CSPs, often those on the receiving end of lopsided handoff ratios, have maintained that the traffic imbalances no longer represent mutual value and have sought to negotiate paid peering arrangements. Their interconnection partners have balked, arguing that each CSP should be compensated by its own end-customer base, and there is no difference in actual cost resulting from the direction in which traffic is carried.
Three Interconnection Issues
As part of this project, the World Economic Forum’s working group on digital infrastructure has facilitated many cross-industry discussions on the increasing contentions surrounding IP interconnection agreements. Three basic issues are at stake.
The first is whether IP Interconnection should continue to be based on commercial agreements or directly regulated. CSPs and others argue that these agreements should continue to be commercial in nature, governed by competition laws. Content providers agree on commerciality, but some are concerned over how CSPs are able to control traffic over the last mile, the only path through which the content provider can deliver traffic to end-users. As content providers see it, CSPs should be required to provide high-capacity connections, since they are compensated by their end-user customers. If CSPs do not adequately maintain their networks on their own, regulators may be required to step in. Most CSPs see only fair business practice in asking content providers to pay for upgraded capacity connections when they send a disproportionate volume of traffic in the direction of the CSP.
Manuel Kohnstamm, Senior Vice-President and Chief Policy Officer, Liberty Global, Netherlands, leads a breakout discussion on IP Interconnection at the World Economic Forum Annual Meeting 2014.
The second issue is whether content companies should help fund network infrastructure upgrades. A number of CSPs say that two-sided commercial models, in which both end-users and content providers compensate the CSP, are required to respond to the infrastructure-related costs of traffic growth and new usage patterns. They contend that if end-customers alone are required to cover these increasing costs, some could be priced out of high-speed internet access. In addition, in their view, the two-sided model imposes an important incentive on content providers to efficiently deliver content across access networks.
Some content providers disagree. In their opinion, CSPs already benefit greatly from the growth in internet traffic as more and better content attracts more customers and leads existing customers to upgrade their broadband connections to plans with higher capacities and prices. Making wholesale carriers or content providers pay to support CSPs’ infrastructure requirements – in addition to their own infrastructure costs – detracts from these companies’ research and development and capital expenditure programmes and could also cause them to pass on the interconnection charges to consumers. Some also argue that paid peering will lead to consumers losing access to some non-commercial content (such as free education and non-profit content) from providers who cannot afford to pay interconnection charges. Both sides have expressed willingness to further explore new business models that make efficient use of the network and promote the development of new digital services.
A third key question is whether different quality of service tiers will lead to better delivery of content, especially high bandwidth and quality-sensitive content, such as real-time video. As the internet has evolved, the content it carries has become more varied. Most content used to be similar in volume and urgency. Today, transmissions can include anything from basic emails to feature-length movies to high-urgency medical images. Many CSPs argue that different users put different demands on the network and that CSPs need flexibility to determine how to transmit data, and at what cost, to manage their networks most efficiently. Establishing different quality of service tiers – one for high-quality delivery, another for “best-effort” delivery (equivalent to how most traffic travels today) – is one proposed solution. Content providers fear that a multi-tiered internet will lead CSPs to favour their own high quality of service infrastructures and potentially underinvest in the upkeep and operation of other service tiers. They point to situations where they claim that CSPs are degrading best-effort quality by failing to augment interconnection capacity. CSPs maintain that high-quality delivery would not negatively affect best-effort quality since they cannot afford to risk losing customers to other CSPs because of poor quality.
Finding Common Ground – On an Uncertain Playing Field
Because of the central importance of IP interconnection agreements to the functioning of the internet, the disagreements over how they operate require attention. They also need a resolution that the disputing parties can call fair and that does not impede the internet’s dynamic growth. A good place to start may be a set of basic insights that can help address these issues through establishing common ground and understanding on the nature of issues under dispute.
The first is the need for common metrics that can provide a fair and accurate assessment of the size of a particular dispute. Although traffic ratios have been used as a core metric for measuring equivalent value in some IP Interconnection exchanges, some parties argue that these ratios may no longer serve this function effectively. They point to other, more pertinent, in their view, means of measurement. Measuring bit-miles (the average number of gigabytes carried multiplied by the number of miles the data travel) has been suggested as an alternative, most notably by wholesale providers; other options include the total transmission costs borne by each network and the value exchanged at interconnection points. No industry consensus has yet emerged on this need, and players on all sides need to continue to work towards an agreement.
Second, while it is customary to think of infrastructure investment primarily in terms of wires and cables, in reality ICT “investment” is much more broad-based. While telecommunications players build the vast majority of network infrastructure, other players invest large sums in vast data centres and cloud infrastructure. There is also a broader universe of companies that invest in creating the digital products – hardware, software and devices, for example – and the services that run on them. Other companies still ensure that the services are effectively delivered and charged for. The total amount of worldwide investment made in “the network” on the one hand and “everything else” on the other is roughly equal.20 (See Figure 7.)
Figure 7: Total Investment in All Other Areas of ICT Value Chain Similar to Telco Capex
Note: Basic methodology leverages market research to estimate market sizes and public company financials to estimate Capex and R&D; assumes capital investments and R&D are both investments in ICT value chain
Source: Telecoms equipment (Gartner), telecoms SW (IDC), devices (IDC), cloud (public company financials), telecoms services (IDC), digital services (BCG), venture capital (OECD); Thomson Reuters Datastream, BCG analysis
Third, differentiating between services that require less or more bandwidth does not constitute discrimination. Varying service tiers may be one of the most useful ways to ensure adequate quality of service for all kinds of traffic as traffic volumes expand. Quality of service differentiation also can open development of new business models, such as multilateral agreements among network operators and content providers for the provision of managed or specialized services. Implementing agreements based on tiered quality of service, however, will require reassurances to some that tiered systems will not be used to avoid or sidestep investment in infrastructure to support lower quality service tiers.
While consumers are not party to IP interconnection agreements, most network participants would agree that it should be easier for consumers to access and understand basic information on the speed and performance of networks, hosting services and digital services. The level of transparency needs to be appropriately calibrated for the intended audience, with the goals of consumers being able to use simple tools to see (to the extent they wish to) where content is coming from, how it is travelling and, in the event of a problem, where – and under whose supervision – it occurred.
A Path Forward?
Despite the uncertainty, discussions on these issues among industry participants have established several important areas of agreement, and these provide a direction for resolving many IP Interconnection issues. The areas of agreement include:
- IP interconnection should remain commercial in nature, as a means for the industry and its participants to keep pace with the pace of change. All parties should commit to avoiding anti-competitive actions and unreasonable discrimination against different kinds of traffic. Industry participants should work to resolve their own disputes; if they are successful the debate over the involvement of regulators is rendered moot.
- Agreements should allow experimentation with new pricing models, again with the condition that the same rules must apply to all players. This experimentation may generate models under which CSPs charge content providers – provided that all parties, including CSPs’ own content operations, face similar types of pricing that do not amount to unreasonable or anti-competitive practices.
- Building private and expedited disputed resolution mechanisms, such as arbitration provisions, into agreements may speed up resolutions when disputes arise. In this type of a commercial environment, disputes are inevitable. They become particularly problematic when they reach the point where consumers are affected. Private arbitration can be written into contracts – assuming both parties agree – as a means to expedite the resolution of these agreements and limit impact on consumers. When the issues under dispute transcend a particular agreement (or arbitration is not provided for), the relevant authorities may need to be called on to help reach resolution.
- Appropriate transparency on network and digital service performance is needed, including last-mile performance even though this is not a direct IP interconnection issue. Forming a voluntary cross-industry body to develop the specific metrics and approaches to measurement that can provide a complete (and unbiased) picture is recommended. A multistakeholder approach, involving a range of participants with clearly differing points of view, has a better chance than either a unilateral approach or a regulated one to agree on a simple set of metrics that adequately captures performance levels of services across the internet.
Sidebar: IP Interconnection and Net Neutrality
IP interconnection agreements affect how traffic is routed among the internet’s wholesale networks, and net neutrality rules apply specifically to the last-mile connection to the end-user’s home or place of business. Definitions of net neutrality are (often hotly) debated, but the essential principle is that all consumer traffic must be treated equally, subject to reasonable network management and network security principles, with potential exceptions for specialized services. What should constitute a specialized service – examples could include medical imaging, video conferencing or even streaming video – is still up for debate. The basic tenets of net neutrality include no blocking or throttling of traffic; appropriate transparency (principally for consumers, but also for content providers) in how traffic is managed within networks; and no unreasonable or anti-competitive discrimination among different kinds of traffic from different sources.
The underlying principle of net neutrality, in the eyes of its advocates, is that neither a CSP nor a government nor anyone else should be able to block or discriminate unreasonably – by price, availability, or otherwise – against any content carried over the internet.
Net neutrality has become a hot political and judicial issue. The European Commission has included a net neutrality guarantee in its “Connected Continent: Building a Telecoms Single Market” package of legislative proposals which passed an important milestone in the European Parliament in April. In January 2014, a US appeals court partially struck down the Federal Communications Commission’s (FCC) 2010 Open Internet Order, which had established net neutrality regulations for CSPs. The FCC has announced that it will issue new rules to “ensure that the internet remains open and fair”.21 Comcast, the largest US cable provider that would become even larger with its proposed acquisition of Time Warner Cable, has already declared that it will continue to abide by the FCC’s long-standing net neutrality principles, regardless of the court’s remand of the specific regulations in the Open Internet Order. Some observers expect little impact; others see a major shift in the internet’s basic business model. The latter predict that bundling various kinds of content for delivery over the last mile, according to an array of pricing models offered by both CSPs and content providers, could become the new normal.
In theory, the court’s decision, and the FCC’s new rules, should have no effect on IP interconnection agreements. Inevitably, perhaps, some issues blur the lines between the two concepts. Quality of service plans are one example. Sponsored data plans are another. Proponents of net neutrality view such plans as open doors for the unequal treatment of traffic and a threat to smaller, innovative start-ups that play an essential role in the internet ecosystem. Those on the other side argue that these plans are not only pro-competitive, they are pro-consumer in that they facilitate the desired content getting into the hands of the end-user who wants it in an economical and technically efficient manner.