Peering, under the hood

The FCC’s initial findings with respect to network congestion and peering arrangements more or less confirmed Verizon’s version of events in its interconnection dispute with Netflix. Whether that will help Verizon and other ISPs as the commission takes up the question of whether to regulate those arrangements is another question.

The FCC this week released its annual broadband performance report, in which it measures average upload, download and streaming speeds under varying conditions across a range of last-mile ISP networks. Overall, the report card was pretty good for the ISPs, with most operators delivering at least 90 percent of their advertised download speeds on average during peak periods. Cable and fiber-based networks, in fact, slightly exceeded their advertised speeds on average, while DSL lines lagged behind (complete data are available here, and in chart form here).

While congestion issues were not a primary focus of the study, the FCC did collect data on traffic flows as part of its speed tests and included those data in the raw data sets it released to the public in conjunction with the report. Here’s how Julie Knapp and Walter Johnston of the Office of Engineering & Technology described the agency’s findings in a blog post:

While internet traffic exchange issues were not the focus of the report, we found notable anomalies as we collected speed data. In particular, we found significant drops in broadband performance during a period when Cogent, an Internet backbone provider, reportedly was having disputes with various ISPs.

Unfortunately, our speed test for the report only measures our sample of consumers connecting to a few other points on the Internet, and it cannot accurately gauge the scope of a network congestion problem. As far as we can tell, this problem only appeared in tests using Cogent links. Furthermore, our research indicates that the issue was “path specific,” i.e., it only affected users when they tried to access certain content or data over certain network paths. However, this problem was still quite significant for those consumers affected.

Cogent, of course, is the transit provider Netflix was using until recently to deliver its traffic to Comcast and Verizon — two of the “various ISPs” with whom Cogent was “having disputes.” And the FCC’s reference to “path specific,” congestion problems comports quite closely with Verizon’s version of what was going on: Netflix chose specific routes, via Congent, by which to send its traffic to the ISPs that it knew, or should have known, were likely to be congested, when other, less-congested options were available (albeit likely at a higher cost to Netflix). It was within Netflix’s power at any time to alleviate whatever performance problems its subscribers were experiencing simply by choosing different routing. In Verizon’s telling, Netflix made its own bed and was simply complaining about having to lie in it.

Separately, data released this week by researchers at MIT also suggested a Netflix-specific problem. The researchers generally found that network congestion was not a major problem in the U.S., but where it did crop up, Netflix was usually in the vicinity.

Whether being right as to the technical issue means Verizon is winning the argument, however, is another matter. That fact that network congestion does not seem to be a major problem except when Netflix is involved — a major competitor for Comcast’s and Verizon’s pay-TV customers — is certainly going to raise red flags with FCC (and perhaps antitrust authorities). Netflix may have had it in its power to relieve the situation but so, too, did the ISPs, as the rapid improvement in performance Netflix subscribers saw on Comcast once Netflix agreed to pay for a direct connection to Comcast’s network made painfully obvious (a similar deal with Verizon has not yet been fully implemented).

“These experiences confirm that the way ISPs exchange traffic with other networks on the Internet is critical to the consumer Internet experience. We need to understand better what is going on and what exactly these issues mean for consumers,” Knapp and Johnston wrote in their blog post. “It’s time for us at the FCC to take a look under the hood and learn whether Internet traffic exchange issues are putting up a barrier to continued innovation and investment on the Internet.”

Netflix doesn’t have entirely clean hands here, either. It clearly was manipulating its own traffic to put Comcast and Verizon in a bad light at the expense of its own subscribers. But it doesn’t need clean hands in this fight. As a political matter, it makes for a much more compelling victim even if it has a hand in its own victimization.