Paid Peering, Paid Prioritization, and the Nuance of the Net Neutrality Debate

It is easy to conflate peering and paid prioritization. Recently, Sen. Patrick Leahy (D-VT) and Rep. Doris Matsui (D-CT) introduced bills in both chambers of Congress to ban so-called paid prioritization. Coverage by the New York Times pointed out that the bills ban “deals similar to the recent agreement that allows Netflix to connect directly to Comcast’s system to avoid network congestion.” However, that is not strictly true. The proposed Online Competition and Consumer Choice Act of 2014 (S. 2476 and H.R. 4880) prohibits “paid prioritization” as understood in the Federal Communications Commission’s conventional discourse on net neutrality. The bills do not explicitly deal with interconnection, or peering arrangements such as the Netflix-Comcast deal. Similarly, in its Open Internet rules, the FCC has so far focused on regulating the potential harms of last-mile paid prioritization rather than that of paid peering/ interconnection. As the FCC reviews the recent flood of net neutrality comments, regulators should be mindful of rhetoric-masked, bad arguments that overlook the nuances between the two.

Paid Prioritization

What is paid prioritization and why are there efforts to ban it? Paid prioritization is a financial agreement in which a company that provides content, services, and applications over the Internet (an “edge provider”) pays a broadband provider to essentially jump the queue at congested nodes. Paid prioritization also covers the cases of broadband providers prioritizing their own content or that of an affiliate over the data from a competing edge provider (also called “vertical prioritization”). What might paid prioritization look like? One example, raised by FCC Chairman Tom Wheeler, is a broadband provider that also owns a sports network giving a commercial advantage to its own network over another competitive sports network trying to reach online viewers. In Verizon v. FCC (2014), the challenge to the FCC’s 2010 Open Internet rules, the D.C. Court of Appeals hypothesized that “Comcast might limit its end-user subscribers’ ability to access the New York Times website if it wanted to spike traffic to its own news website, or it might degrade the quality of the connection to a search website like Bing if a competitor like Google paid for prioritized access.”[1]

Why might paid prioritization be harmful? The routing of Internet Protocol (IP) data is a zero-sum game. With finite bandwidth capabilities, the creation of a “fast lane” by an Internet service provider (ISP) entails the implicit creation of an accompanying “slow lane” for other bytes of data not being sped up. Net neutrality advocates oppose Internet fast lanes on the grounds that only a limited and exclusive group of providers will be able to pay for priority, resulting in sufficient harms to warrant FCC regulation. These harms include hindering innovation, impeding competition, and undermining consumer rights. For instance, a fledging startup with a superior product may still be drowned out if its online services are slower than an established competitor who can afford paid prioritization. A small-town broadcasting station trying to reach online listeners may lose out to big stations which can strike deals with ISPs so their content does not count against mobile users' wireless data caps. In other words, on an unequal playing field, products and technologies do not rise and fall on their own merit.

The FCC would like to ban paid prioritization, but is seeking the right legal authority to ground its regulations. The FCC’s two previous attempts to enact Open Internet rules have been struck down by the same federal appeals court, first in Comcast v. FCC (2010) and again in Verizon v. FCC (2014). In April 2010, the D.C. Circuit Court of Appeals ruled that the FCC does not have ancillary authority over Comcast’s Internet service. In January 2014, the same court ruled that the authority granted to the FCC in Section 706(a) of the Telecommunications Act of 1996— “[encouraging] the deployment on a reasonable and timely basis of advanced telecommunications capability to all Americans” — is insufficient to enact per se common carriage regulations over the Internet. Subsequently, the Notice of Proposed Rulemaking (NPRM) released by the FCC in May 2014 acknowledges that Section 706(a) “could not be used” to prohibit pay-for-priority practices. As the FCC seeks alternative legal authorities to ground its regulations – be it the “commercially reasonable” standard, the more strenuous Title II, or others[2] – Chairman Wheeler has emphasized that “providing exclusive, prioritized service to an affiliate is not commercially reasonable.”

Paid Prioritization vs Peering: The FCC's View

As the FCC’s legal basis for regulation evolves, its approach in treating paid prioritization versus interconnection has so far stayed fairly constant – i.e. attempting to regulate the former but not the latter. After Comcast v. FCC (2010), the Commission, led by former Chairman Genachowski, passed an Open Internet Order in December 2010. The 2010 Order states that the FCC did not intend its rules prohibiting access fees under the “no blocking” principle to “affect existing arrangements for network interconnection, including existing paid peering arrangements.” Spelt out more explicitly in the FCC’s own words, the no-blocking or unreasonable discrimination rules did not apply beyond “the limits of a broadband provider’s control over the transmission of data to or from its broadband customers.” The same intention to regulate paid prioritization but not paid peering has persisted since then. In May 2014, the FCC tentatively concluded in the NPRM that it would “maintain this approach” of separating peering/interconnection from paid prioritization, but sought comments on potential changes. Specifically, the Commission asked whether to continue its approach “not [to] apply the no-blocking or unreasonable discrimination rules to the exchange of traffic between networks.” In response, the Benton Foundation filed comments jointly with Public Knowledge and Access Sonoma Broadband to the FCC in July 2014, stating that addressing interconnection is essential to protecting the Open Internet both for residential customers and edge providers.

The Straw-man Argument that "the Internet is Inherently Unequal"

A common rebuttal to the harms of paid prioritization is that “the Internet is already unequal.” This camp posits that the Internet is already not an utterly level playing field in the status quo, hence there is little value in prohibiting paid prioritization. They point out that discrimination of data is already rampant via institutionalized peering and interconnection agreements.

True, there are other ways that edge providers can pay to obtain advantages in the delivery of their content. These alternative efforts aim to increase performance at the backbone level. They exist beyond the last-mile between the ISP and the end-user. Examples of such inter-network transmission of data include content delivery networks (CDNs), settlement-free peering, paid peering, and provider-owned facilities that are dedicated solely to interconnection.

Those who argue that “the network is already not neutral, and therefore regulating paid prioritization is superfluous” are conflating the local access network and the Internet backbone. There are differences both in the extent of competition and the nature of data differentiation between the two data delivery components, as explained below. While paid prioritization conventionally refers to content discrimination in the last-mile that do fall under FCC’s net neutrality rules, peering arrangements have not been regulated by the FCC, in part because of more vibrant competition in the Internet backbone.

What is Peering and How does it Occur?

How do peering and paid prioritization fit into the Internet? While most people today are aware that the Internet is more than “a series of tubes,” any satisfactory answer involves delving into the Internet infrastructure, which is complicated to say the least. Caveating that any answer in the form of a blog post does not do it complete justice, here is a simplified version with just enough information to differentiate between peering and paid prioritization. For more granular explanations of how peering fits into the Internet infrastructure, refer to the work by Dr. Anna-Maria Kovacs, Visiting Senior Policy Scholar at the Center for Business and Public Policy at Georgetown University, or the resources from the Global Peering Forum in 2009 assembled by William Norton, an expert on peering and transit agreements.

Let us start with Tier-1 network providers. They are commonly defined as networks that can reach/interconnect with anywhere in the specific region of the global Internet backbone without paying settlements or buying IP transit (hence “transit-free”). They have the highest position in the data delivery pyramid. Tier-1 providers constitute a small club, according to the ”Baker’s Dozen” 2013 report by renesys, with the top five being, in this order, Level 3, TeliaSonera, NTT, GTT, and Cogent. These Tier-1 backbone providers buy or build connectivity.

Lower-tier ISPs (also called “regional” or “Tier-2” ISPs) pay transit fees to connect to Tier-1 ISPs.[3] Transit fees payers can then reach the entire Internet. Now, “peering” enters the picture. Peering is when two or more networks interconnect directly with each other for traffic exchange. Bill Norton calls it a “business relationship by which two companies reciprocally provide access to each other’s customers.”

Traditionally, at the early stages of the Internet, all peering agreements involved no settlement costs (“settlement-free peering”) and were conducted between Tier-1 ISPs. Since then, peering entities have evolved to also encompass Tier-2 ISPs, CDNs, and even edge providers. Peering deals can also involve payment (“paid peering”), often if there is traffic imbalance between the negotiating parties. The price is usually determined by the level of mutual benefit. How that price is determined, and the choice between paid or settlement-free peering, has been the point of contention between network providers. If an ISP faces extra traffic that its own network cannot handle, the first choice to solve its congestion is usually to find suitable networks to peer with for free, since paying for IP transit, paid peering, or upgrading infrastructure etc are more expensive alternatives. ISPs, like Comcast and Level 3, publish their own peering policies. They have different preferences regarding how much to rely on settlement-free versus paid peering. Peering practices demonstrate clearly how the Internet backbone is what economists call a “two-sided market,” or more aptly, an “n-sided market,” – where both negotiating parties provide each other with mutual network benefits. Known examples of peering agreements, some settlement-free and some paid, include those between Tier-1 providers (e.g. Level 3 and Cogent, Level 3 and XO Communications), between Tier-1 and Tier-2 providers (e.g. Comcast and Level 3, Verizon and Cogent, Level 3 and Verizon), between Tier-2 providers and edge providers (e.g. Comcast and Google), and many others.[4]

Finally, end-users pay Tier-2 ISPs for broadband access, connecting the “last-mile”, the connection between the end-user and the ISP. It is at this very last stage where the mass-market ISP may act as a “gatekeeper” to end-users. Even “multi-homed” end-users – those who have access to more than one ISP – cannot switch easily without incurring high transaction costs. In short, since the last-mile traffic within your local access network has to go through your ISP to reach you, cable or telephone companies often act as “terminating access monopolies” – where the single provider has leeway to profit-maximize in ways often unfriendly to consumers. The last-mile is the part of the process where the FCC has tried to prohibit paid prioritization, which the Commission views as commercially unreasonable.

Implications on Protecting and Promoting an Open Internet

In contrast to the last-mile, the Internet backbone (i.e. between the Tier-1 provider and the interconnection point of the last-mile ISP) presents a more competitive market. For example, edge providers can purchase CDNs from companies like Akamai, Cloudflare, and Limelight.[5] Companies who can afford it can even build their own CDNs (e.g. Google, Microsoft, and Yahoo). They can also directly purchase better connectivity (“bigger pipes”) from Tier-1 network providers like Level 3. A startup edge provider trying to improve its content delivery quality and speed is not currently starved for options. These non-last-mile inter-network arrangements – transit, paid peering, settlement-free peering, CDNs, etc – are not equal to, and should not be confused with, last-mile paid prioritization. Clumping “paid prioritization” and “interconnection/paid peering” together in the discourse detracts from the merits of the arguments. This “straw-man” tactic cheapens the debate, regardless of if you believe the Internet should be regulated with a light or heavy touch.

Interconnection/paid peering is believed to enhanced existing infrastructure, though the jury is out on potential spillover effects, hence regulators ought not over-generalize the effects of peering agreements on net neutrality. Mozilla argued in its comments that interconnection deals create additive gains to network capacity and efficiency, unlike the zero-sum game of paid prioritization. The details and potential effects of nascent peering agreements are awaiting scrutiny. Netflix, the world’s biggest on-demand Internet TV and movie streaming service, entered into two peering arrangements in 2014: first with Comcast in February and then with Verizon in April. The agreements were among the first of its kind. They cover interconnection and do not fall under existing FCC net neutrality rules. There may well be effects on network openness which ripples we have yet to see. Following the Comcast deal, Netflix CEO Reed Hasting complained that his company is “reluctantly” paying an “Internet toll.” While video streaming providers like Netflix and Youtube are publicly shaming ISPs, ISPs have hit back by placing the blame on content providers. For example, Verizon pointed out that Netflix could cut buffering by making “simple adjustments.”

Chairman Wheeler would like to ban paid prioritization and is gathering more information on peering before taking a stance. The Chairman has indicated that paid prioritization is not “commercially unreasonable.” Commercial reasonableness is one of the laxer Open Internet standards currently being considered by the FCC. The Commission is still trying to parse through known peering arrangements. In June 2014, the FCC opened an inquiry into interconnection deals to gather more data. Chairman Wheeler has stressed that “what [they] are doing right now is collecting information, not regulating.”

To achieve its goals of protecting innovation, competition, and consumer rights, the FCC will inevitably need to address interconnection. Ars Technica recently reported that despite the peering agreement, Netflix performance on Verizon FiOS has dropped. Despite both Comcast and Verizon reaching peering deals with Netflix, subscribers received very different results. The incongruence demonstrates potential problems with lack of transparency in peering deals, as well as tangible harms to a vast number of consumers. Chairman Wheeler has repeatedly indicated that “consumers must get what they pay for.” Opaque peering agreements make it more difficult and frustrating for consumers to hold their providers accountable. ISP gatekeeper concerns that have traditionally animated Open Internet discussions are very much alive and relevant in the peering context. Paid prioritization and interconnection are key pieces of the net neutrality puzzle. As the latter continues to get more attention and coverage, regulators must craft policies that will uphold the Commission’s core values when applied to both the Internet backbone and the last-mile.

Notes:

1. These are explicit examples that demonstrate why content-gatekeeping by ISPs may disadvantage smaller actors in the market. Similar concerns have also been raised about more subtle practices like "zero-rating" – exempting certain apps or content from a data cap – in mobile broadband.
2. The Benton Foundation has publicly supported the reclassification of broadband access service under Title II since 2010.
3. “Transit” is when a network agrees to carry the traffic that flows between another network and all other networks, in return for a transit fee. Also note that ISP tiers do not necessarily indicate enterprise market share. The TeleGeography Global Internet Geography Report 2013 found that “while Level 3 was the largest backbone provider in the world in terms of overall number of connections, IP address share, and reach, the company ranked third behind Verizon and AT&T in terms of Fortune 500 customers […] As with Verizon and AT&T, former PTTs and other old, well-entrenched carriers tend to have a stronger presence in the Fortune 500 market than in the Internet sector, overall.”
4. Note that these labels are not absolute and many network entities perform multiple roles. E.g. Level 3 enters into many settlement-free peering arrangements and also provides CDNs. Google is both an edge provider and runs a CDN.
5. CDNs utilize the rationale of local caching. Data stored in closer proximity to end-users get delivered quicker.

Raphael Leung is an intern at the Benton Foundation. He is a senior at Yale University and an Undergraduate Fellow at the Information Society Project at Yale Law School. Contact him at rleung [at] benton [dot] org