New research from computer scientists at the Massachusetts Institute of Technology and the Center for Applied Internet Data Analysis shows that, as its traffic increased in the last two years, Netflix forced it through clogged delivery routes, when less congested channels were available.
The scientists concluded the data “strongly suggest that the correct response to growing congestion” was for Netflix to use different connections for its traffic, not for Comcast to add capacity. This corroborates Comcast’s argument that “Netflix has not been honest.”
Evidence also suggests that, at the same time Netflix allowed congestion to increase until it affected performance, Netflix published misleading data casting the blame on Comcast.
This one-two punch prompted Netflix subscribers to pressure the Federal Communications Commission to adopt policies supposedly designed to reduce congestion but that were actually aimed at lowering Netflix’s costs of doing business. In other words, Netflix first gamed the network, and then the policymakers themselves.
Here’s how Netflix did it. In June 2012, Netflix launched a “content delivery network” or “CDN” service to stream video to Internet providers. CDNs are basically middlemen that specialize in carrying large volumes of traffic across the backbone of the Internet. Netflix said it hoped to shift most of its traffic from its existing routes to its new CDN.
But Netflix wasn’t telling the whole story. It recently admitted that in early 2012 it also began shifting its traffic onto what are called “settlement free” routes for delivery to Comcast’s network. These “settlement free” routes were cheaper for Netflix, but lacked flexibility and capacity. Netflix knew this decision, entirely within its own control, was degrading its own subscribers’ service. Predictably, the settlement-free routes Netflix chose were overwhelmed by Netflix’s enormous streaming volume.
According to Comcast and the MIT and CAIDA scientists, Netflix could have eliminated the congestion overnight by shifting traffic to uncongested routes on which Comcast had capacity. This would have solved the service problem, but would have cost Netflix some of the profits it was saving by avoiding commercial CDNs.
Instead of solving the problem, Netflix decided to spin it. Netflix published “ISP speed rankings” to provide “monthly insight into which ISPs deliver the best Netflix experience.” But these rankings were skewed to shift blame off Netflix’s own choices and onto ISPs. In August 2013, for example, Netflix published a regional speed ranking purporting to show average speeds for Netflix on Boston’s RCN network outperformed Comcast and other ISPs as much as 70 percent. Notably, Netflix didn’t mention that it directly connected its CDN to RCN in Boston and avoided the settlement-free routes it used with Comcast.
In February 2014, a few weeks after a federal court overturned the FCC’s Net neutrality rules, and the same month Comcast announced its merger with Time Warner Cable, one Net neutrality advocate leaped to the conclusion that Comcast was “definitely throttling Netflix.” When Netflix was asked to comment on the allegation, it didn’t mention its Internet routing decisions. Instead, Netflix fueled speculation on “throttling” by referring to its ISP speed rankings and claiming that Comcast’s Netflix speeds had been dropping for months.
This polluted the important debate over net neutrality. While Netflix included the disclaimer that its rankings are affected by many factors, it argued that these “cancel out when comparing across ISPs.” But that’s not true when the difference is caused by decisions within Netflix’s control.