Edward Snowden’s NSA surveillance disclosures have drawn global attention to the sometimes esoteric world of Internet infrastructure governance. Political reactions have ranged from wanting to “route around” American Internet exchange points to walling off services through data localization and nation-specific cloud computing. But these responses sometimes overlook how the Internet works in practice.
Now Comcast and Netflix have announced that they will directly interconnect their networks, rather than having Netflix traffic flow first through a third-party network. With this, another layer of Internet architecture—interconnection and peering—is under the microscope. The Internet is not actually a cloud but a collection of networks that technically conjoin, or “interconnect,” and exchange traffic based on negotiated business arrangements known as “peering” or “transit” agreements.
Many have reacted with anger to this recent interconnection announcement, viewing it as a direct threat to the principle of net neutrality. But the responses here again underscore the lack of technical and economic understanding of how the Internet works—which could be attributable to the lack of transparency in this space.
Here are five things to keep in mind when you think about the Netflix-Comcast (and soon, perhaps, Netflix-Verizon and Netflix-AT&T) deal.
1. Content companies are already globally connected to Internet access providers.
Many have claimed that this is the “first time a content company has connected directly to a consumer network or paid for interconnection” and that it will threaten the open Internet. In reality, long gone are the days of a tiered system in which so-called “Tier 3” Internet service providers paid “Tier 2” intermediary networks that, in turn, paid “Tier 1” global backbone providers for access to the global Internet. Interconnection is more organic, messy, and decentralized than that. Large content companies like Google and Facebook distribute and replicate their content all over the world, connecting directly to Internet access provider networks either from their own server facilities, via Internet exchange points (IXPs) or via content delivery networks like Akamai and Limelight.
From an engineering perspective, bringing multimedia content closer to “eyeballs” without clogging up a tertiary network is a no-brainer: It decreases delays and optimizes bandwidth consumption. This content distribution is particularly important for bandwidth-consumptive applications like video.
2. Paid interconnection is already the norm.
The agreement is also not a unique instance of a content company compensating a network operator for interconnection. Numerous types of paid and unpaid connection arrangements are struck, depending on factors such as traffic volume, network reach, and performance guarantees. “Settlement-free peering agreements” involve network operators agreeing to mutually exchange traffic without payment. “Settlement-based interconnection” involves one network paying another network for either mutual peering or for transit to the global Internet.
Why would it be more acceptable for Netflix to pay another company, Cogent, for connectivity to Comcast than for Netflix to pay Comcast directly? If anything, a content company connecting directly to an edge network is in the public interest because it avoids susceptibility to a possible outage from peering disputes between an intermediary network and a customer network. Recall that a 2008 Internet outage occurred because of a Cogent-Sprint interconnection dispute. Netflix connecting directly to Comcast circumvents possible outages due to peering disputes beyond its control.
3. This is not a death knell for net neutrality.
The net neutrality issue of whether network operators should be allowed to discriminate (throttle back or prioritize) certain content over other content isseparate from interconnection, and the companies’ press release states, “Netflix receives no preferential network treatment under the multi-year agreement.”
Given the recent ruling on net neutrality and the history of complaints against Comcast for throttling peer-to-peer file sharing traffic, it is not surprising that many conflate interconnection with net neutrality traffic prioritization concerns. This deal would also not be receiving as much attention if there was adequate broadband competition and less media concentration (think proposed Comcast-Time Warner Cable merger). But, to use a physical world traffic example, interconnection is about building new roads and net neutrality is about speed limits and speed bumps on the roads. There is a significant difference between establishing a no-discrimination rule versus regulating who is allowed to connect to whom.
4. There are good public-interest reasons to pay attention to interconnection between companies.
Arrangements of technical architecture are also arrangements of power. Interconnection is a public interest issues because it provides the basic infrastructural foundation for the digital public sphere, because it promotes Internet growth and efficiency, because shared interconnection points can serve as concentrated sites for government censorship and surveillance, and because peering disputes can result in outages. For all of these reasons, more (not less) distributed and decentralized interconnection is a good thing. With existing antitrust law in mind, it is also not realistic to introduce government regulations mandating who can connect to whom, how, and under what terms.
5. Interconnection needs more transparency and best practices.
What would be useful, however, and what might have pre-empted current confusion about interconnection, is network operator transparency about the terms of interconnection agreements and a set of industry best practices for these agreements. Similar to transparency reports for content intermediaries, this would help bring the hidden world of Internet infrastructure into the light.
Maybe we can read these reports while watching new Netflix episodes ofOrange Is the New Black.