Innovation Files has moved! For ITIF's quick takes, quips, and commentary on the latest in tech policy, go to

Netflix Grows Up

Last month something evil happened in Las Vegas: Netflix was invited into the inner cloister of the Last Mile Cabal, where a blood sacrifice sealed a dark pact with Comcast. What was in that pact, what were the terms on which the sacrifice was made? I’ll tell you: a commercial transaction that will reduce congestion at points of interconnection, improving Netflix performance across Comcast’s network, bringing joy and good cheer to video streamers across the country. Wait, what – you may ask – what’s evil about that? I’m not quite sure either, although reading the coverage of this deal, you’d think it was.

To be clear, this is an interconnection issue, not a net neutrality issue.  Let me repeat that: this is not a net neutrality issue.  It is unfortunate timing for the parties – with the recent opinion from the D.C. Circuit vacating the Commission’s non-discrimination and no-blocking rules some industry watchers are on hair-trigger to find a would-be violation. Accusations that Comcast was “throttling” Netflix, or that Netflix is “paying off” Comcast for a “premium” connection are simply wrong. Netflix is not getting priority treatment of its traffic, but simply routing around an under-performing middle-man to ensure that its massive amount of data makes it to its viewers with high quality. This is certainly not the first time a content distributor has paid for interconnection, nor is it unusual or particularly troubling – except for the middleman who lost business!

Many Innovation Files readers probably already saw Dan Rayburn’s excellent clarification of the issue, but if not, it is worth checking out. Also see his latest update on the Netflix/Comcast speculations. Dan’s clarification is commendable in a world where technical facts very quickly drop out of telecom policy debates. I must confess guilt of being lazy on the speed/throughput distinction in discussing broadband , but Dan is right, unless you are a high-frequency trader, you probably don’t care about your broadband speed. Confusing speed and throughput is relatively innocuous, but once these distortions pile up, we quickly have people thinking this interconnection deal implicates net neutrality. Some are quick to assume the worst of cable companies, and this bias can quickly amplify when fueled by distorted simplifications. Take, for example, an analogy Tim Wu opens with in an op-ed discussing the Comcast – Netflix deal in the New Yorker last week.

Imagine a restaurant whose most popular dishes are made with fresh tomatoes. Given the large quantity in which these dishes are produced, tomato shortages are frequent, yielding patron complaints. In response, the restaurant’s tomato supplier promises to build a storage facility nearby, in order to insure that the restaurant never runs out of tomatoes.

In a competitive market, the restaurant would gladly accept this offer, fearing that its uneven tomato supply will cause it to lose customers. But, if the restaurant industry resembled today’s broadband-Internet market, the restaurant would be the only place in town that served tomato dishes. Facing no competition, it might try to extort extra payments from the tomato supplier, knowing that without its business the supplier would go under.

 Let’s unpack this analogy: the restaurant is presumably Comcast, and the tomato dishes presumably Netflix. The economics of restaurants and broadband networks are very different, so adjusting the analogy to reflect reality may strain our understanding of normal restaurant operations, but we should value accuracy over rhetorical appeal on important policy questions. Wu claims that “if the restaurant industry resembled today’s broadband-Internet market, the restaurant would be the only place in town that served tomato dishes.” This seems confused at best.

While the consumer broadband market is far from perfect competition (which, as an aside, is a good thing – excessive fragmentation in network reduces incentive and ability to invest in their network), cable is still far from a monopoly. Telco companies provide a very similar triple-play package, with competitive or higher speeds, excuse me, throughput, than cable. Satellite operators also compete, especially in the video market – their broadband offerings suffer from slower data rates and higher latency. Furthermore, Netflix is far from the only video content available. The over-the-top market is competitive, as is the traditional satellite, cable, telco offerings. The particular economics of facilities based competition in networks providing access to an IP platform with increasing returns to scale means there is no clear analog to restaurants, but cable is far from the only restaurant in town serving tomato dishes, whether we mean video generally or Netflix specifically.

Let’s give Wu the most charitable reading we can and consider Netflix as the dishes and the data stream as the tomatoes. Even when we grant Comcast as one of three or four restaurants serving Netflix for dinner, this deal still doesn’t seem unreasonable. Netflix is responsible for a little over 30% of peak downstream traffic in North America – imagine if 30% of traffic coming in through our restaurant’s front door were tomatoes. This deal is akin to Netflix using the freight delivery in the back instead of the front door.

Some are concerned that Comcast and all the other major last-mile providers have declined to use Netflix’s Open Connect system, an appliance that caches the streaming service’s most popular content closer to the consumer. It makes sense that Netflix would go this route, working to ensure their consumers have the best viewing experience. Both the Open Connect appliance and the recent interconnection deal cut out the middle-man transit provider to better align incentives. Although some content delivery networks will pay to provide the interconnection throughput their clients need, other transit providers are interested in keeping their costs as low as possible, regardless of the impact on clients’ service.

Both Open Connect and direct interconnection are a positive development for consumers, allowing Netflix to ensure a certain level of quality – the question is whether the last mile network is justified in seeking compensation. Contrary to Wu’s assertion that “[o]nce a cable company’s infrastructure is in place, it costs it almost nothing to provide actual service,” there are real, ongoing costs to running a broadband network. There are even considerable operating costs to simply running the Open Connect appliance itself, such as power, space, cooling, servicing, etc. Much of the last mile network has been upgraded to provide the throughput necessary for streaming high-definition video – those are real costs that must be recovered somehow. I suggest the one responsible for 30% of peak traffic is a reasonable place to start.

Some have suggested that this will bolster broadband operators to seek payments from other video upstarts, preventing innovation and locking out new entrants. This is very unlikely.  Start-ups will continue to have the same opportunity to grow and prosper using the same transit options to connect to the wider Internet – it’s only when they grow to be the size of Netflix (or perhaps a little smaller) that interconnection starts to make sense. Considering the economics of settlement-free peering, how Netflix’s transit providers were connecting with Comcast prior to the deal, and why it arose in the first place indicates that it isn’t going anywhere anytime soon.

Settlement-free peering is used because it greatly reduces transaction costs. Frequently these agreements are astonishingly informal, made over a handshake, allowing easy and quick connections between networks. The requirements for peering generally include a rough balance in traffic flows at a set number of interconnection points. The last-mile networks have little interest in tracking down every single new video provider to draw up and negotiate over a contract. Once a new entrant gets large enough to contribute a substantial amount of traffic and starts imposing real costs on the network, then custom interconnection arrangements make a lot of sense for all parties involved.

Print Friendly, PDF & Email