Business relationships between major broadband providers and Internet backbone providers appear to be the cause of major drags on performance from early 2013 to early this year, according to a new study from a think tank advocating for strong net neutrality rules.
Five major U.S. broadband providers, including Comcast, AT&T and Verizon Communications, saw major performance degradation during those months, according to the study published Tuesday by the New America Foundation's Measurement Lab Consortium [M-Lab].
Broadband providers questioned the study, saying it wasn't able to pinpoint the exact cause of congestion.
For some customers of major broadband providers, download speeds were less than 0.5 Mbps for months at a time, said M-Lab, which used network diagnostic tools installed on Internet backbone providers' networks.
M-Lab found frequent traffic congestion problems when specific broadband providers were connected with specific backbone providers, the study said. "From this we conclude that ISP interconnection has a substantial impact on consumer Internet performance -- sometimes a severely negative impact -- and that business relationships between [broadband and backbone providers], and not major technical problems, are at the root of the problems we observed," the study said.
The study resurrects a contentious debate about whether proposed net neutrality rules at the U.S. Federal Communications Commission should apply to backbone interconnection deals. That debate, dating back to Netflix complaints about Comcast in 2010, has flared up during the past year as the FCC debates new net neutrality rules.
The New America Foundation's Open Technology Institute has advocated for strong net neutrality regulations. The foundation gets funding from several sources, including net neutrality supporters Netflix and Google.
Broadband providers disputed the conclusions of the study, however. While the study points to backbone business relationships as the cause of traffic degradation, it isn't able to pinpoint which business relationships were the problem, said David Young, vice president of public policy at Verizon.
Other studies have pointed to Netflix business deals as the cause of traffic degradation, broadband providers said. "This is a problem that's been discussed pretty extensively and documented pretty extensively," Young said.
AT&T didn't mention Netflix by name, but said commercially negotiated traffic peering agreements on the Internet need to be balanced.
"When one side of a commercial peering arrangement sends significantly more traffic than it receives, the allocation of infrastructure costs described above gets skewed," Bob Quinn, AT&T's senior vice president for federal regulatory policy, said in a statement. "If the sending party refuses to take steps to bring balance back into the arrangement, congestion can result."
When severe congestion happens, it "can only be addressed by a new commercial arrangement involving steps to either reduce the traffic flows, route the traffic in a more efficient way, invest to augment capacity, or some combination of all three," he added. "This is how the Internet has always worked -- and it works very well."
The M-Lab study found significant congestion on some broadband providers' networks between early 2013 and early this year. In New York City, customers of Comcast, Time Warner Cable and Verizon saw drops in download speeds, in some cases down from 20 Mbps to 4 Mbps for extended periods.
But broadband provider Cablevison, which like the other three providers also uses Cogent's backbone to provide service, did not see similar congestion, suggesting that Cogent "had sufficient capacity" in at least part of its network, the study said. The traffic problems in New York City and some other cities abated in February at all the broadband providers experiencing problems, the study noted.
There were similar patterns in other cities. In Dallas and Los Angeles, customers of five large broadband providers saw massive congestion between February 2013 and February of this year, while customers of Cox saw no significant congestion, the study said.
Vint Cerf, a Google vice president and Internet pioneer, praised the M-Lab study, saying it suggests that business relationships between broadband and backbone providers were the source of the severe congestion.
The M-Lab study is "the first work of its kind, using open data and reproducible methods to expose complex performance issues at scale," Cerf said in a statement. "This research highlights the critical role of open data and transparent research in maintaining an accessible and open Internet: Without knowing the scope or extent of the problems, we cannot act to remedy them."
Grant Gross covers technology and telecom policy in the U.S. government for The IDG News Service. Follow Grant on Twitter at GrantGross. Grant's email address is [email protected]