Interconnection: Towards a New Regime
Internet interconnection usually doesn’t make for big news. The term refers to the agreements that connect up the Internet’s component networks and since they usually “just work” they rarely attract the media’s spotlight. We don’t particularly care how our House of Cards gets to our screen or whether our video bits travel long distance though a transit provider or cached closer to home on a content delivery network (CDN), so long as it works. So long as it works. Now old news, it was around October when some Netflix users first began seeing their streams slow.
This touched off a rather public spat, with a back and forth of blog posts from Netflix and Verizon. Tensions rose after Netflix began placing notices on customer’s screens accusing Verizon’s network for the slow streams. Verizon responded with a cease and desist with which Netflix appears to have complied. On top of all this, FCC Chairman Tom Wheeler recently released a statement announcing a relatively informal investigation into recent interconnection deals.
On Wednesday, at an event hosted by the Congressional Internet Caucus Advisory Committee, David Clark, noted Internet engineer and MIT researcher, presented preliminary results from a joint MIT – University of California at San Diego study on the causes and locations of congestion within the core of the Internet. It’s nice to actually have some actual data on an issue instead of speculation and accusation. The researchers were cautious and reiterated that their results are preliminary, but the conclusions of their abstract read as follows:
- Our data does not reveal a widespread congestion problem among the U.S. providers.
- Most congestion we see can be attributed to recognized business issues, such as interconnection disputes involving Netflix. These issues are being resolved, if slowly.
- Congestion does not always arise over time, but can come and go essentially overnight as a result of network reconfiguration and decisions by content providers as to how to route content.
The little press coverage I’ve seen of this event has focused on how little the panel actually discussed the data and how both the broadband ISP industry and Netflix continued to point fingers at one another. True, the panel did quickly devolve into the same arguments: Netflix accuses ISPs of deliberately congesting their points of interconnection and refusing to give them free access to their customers that already pay the ISP for a pipe to Netflix. The ISPs, on the other hand, accuse Netflix of choosing inappropriate interconnection methods, deliberately avoiding the numerous settlement-free points of interconnection available and unfairly invoking the net neutrality hot-button as a way to pull in support for zero-price carriage of their traffic or stronger leverage in interconnection negotiations. See our past writings to get a sense of where ITIF thinks the truth lies.
Nevertheless, I think there are some key take-aways from this new data.
First, it is worth noting the point made in Dr. Clark’s presentation that there are numerous reasons a broadband user could have a frustrating experience with their broadband, and interconnection congestion is only one of them. Although many people’s first impulse is to blame their ISP, this is rarely the correct response. Clark points to WiFi interference as a typical culprit for slow streams, but there are a slew of others. Some congestion exists, but there are many other places things can go wrong, most of which are outside an ISP’s control. The notion that ISPs deliberately slow the data rates available to their customers makes little sense and, by the way, is not at all supported by recent data collected by the FCC that shows ISPs deliver, on average, 101 percent of advertised speeds.
More importantly, congestion in the core of the network is rare and where it does occur it is because of real disparities between capacity and demand and not problems with the technology. In other words, where there are interconnection problems, it is usually because of specific business disagreements. There was some limited congestion on particularly expensive trans-oceanic links, but the main interconnection problems the researchers found were all specific to extremely bandwidth-intensive video streaming services, and not to simple access to websites (e.g., google.com, amazon.com, etc.)
As Clark puts it, “Netflix brings its own congestion.” The amount of bandwidth that Netflix users draw onto last-mile networks is totally unprecedented and requires changes in the way content is delivered. Indeed, it is estimated that in evening hours Netflix accounts for around 1/3 of all Internet traffic. These changes have been ongoing for years, most explicitly in the growth of the CDN market. The interconnection market has flourished without regulatory oversight or price restrictions (let alone a requirement for free interconnection) for decades. If Netflix is in fact having trouble delivering its content (and it is) it should seek out new types of business arrangements. These public business disagreements are not signs that the Internet is broken, or that anyone is unduly exercising market power – these are simply growing pains in one more phase of the Internet’s maturity.
One arrangement that likely makes the most economic sense for a service that uses extreme amounts of bandwidth like Netflix would be to directly interconnect with last-mile networks. Netflix representatives have claimed that they are using five different transit and CDN providers to deliver their traffic (Netflix will often shift from one transit provider to another based on who can offer them the lowest price, but not necessarily the most reliable through-put). Direct interconnection would likely lower their costs and certainly improve their customer’s experience. Note that this is exactly what Netflix did. The company came to agreements with two ISPs, and may well be in talks with others. Only after those contracts were secure did it reverse its position and begin complaining about the supposed “tolls” they were being charged.
The choice of forum for these complaints alone speaks volumes. If Netflix has a legitimate complaint about the deals they got, if it really thinks that the terms of the agreements they voluntarily struck with Verizon and Comcast are somehow anticompetitive, they can go to an anti-trust authority or petition for a formal proceeding with the FCC. Instead, the company chose to launch a PR campaign about “Strong Net Neutrality” through the company blog. This move was clearly designed to whip up the net neutrality crowd which was already foaming at the mouth from the January court decision vacating parts of the 2010 Open Internet Order.
It appears to have had some effect: when Chairman Wheeler announced he was conducting an informal inquiry into the matter, he cited the number of complaints he received from consumers in the Open Internet docket pointing to the problem. I suspect the Chairman will find that nothing particularly evil is happening in these contracts, just as France’s regulator found after a very similar dispute between France Telecom and Cogent.
It should also be noted that hosting high-bandwidth content either within the last mile or a similar direct interconnection is a much more efficient way for it to be distributed than through numerous transit providers. These agreements are something to be celebrated, not concerned over. Direct interconnection can open up previously congested transit ports, allowing other transit customer’s data to flow unimpeded.
Netflix claiming that this particular way of delivering its traffic, for which it was previously paying five different transit providers, should be provided for free is frankly a little absurd and not very helpful to resolving the issue (except, of course, as leverage for their own interests in negotiations). Netflix should be approaching these negotiations like they would any other transit or CDN service they regularly contract for.
We shouldn’t let fears of abstract potential harms stand in the way of normal business transactions and efficient delivery of our favorite video. I’m sorry, but the fact that Netflix is paying someone new for the same service doesn’t ring any alarm bells here.
