May 15, 2008, 2:50 PM — UPDATE: Information in the fourth paragraph saying the tests of Cox's network were limited in number and Cox reaction in the eighth and ninth paragraphs have been added. And the word "first" before "reported" in the first sentence of the 14th paragraph has been removed.
U.S. cable broadband providers Comcast
and Cox Communications are
slowing BitTorrent
traffic at all times of the day, not just during peak traffic, according to
a new study by a German computer research group.
Comcast has insisted that it uses network management techniques to slow some
peer-to-peer traffic during times of peak congestion, but the study from the
Max Planck
Institute for Software Systems suggests that Comcast and Cox are slowing
BitTorrent traffic "independent of the time of day."
The study,
using more than 8,000 nodes worldwide to test for BitTorrent blocking, found
that Comcast was interrupting at least 30 percent of BitTorrent upload attempts
around the clock. At noon, Comcast was interfering with more than 80 percent
of BitTorrent traffic, but it was also slowing more than 60 percent of BitTorrent
traffic at other times, including midnight, 3 a.m. and 8 p.m. Eastern Time in
the U.S., the time zone where Comcast is based, according to tests run by users
of the institute's Glasnost network testing tool.
Cox was interfering with 100 percent of the BitTorrent traffic at 1 a.m., 4
a.m. and 5 a.m., also Eastern Time, according to the tests. However, there were limited numbers of tests of the Cox network, with as few as two users testing during some time periods.
Comcast downplayed the results. P-to-p traffic makes up 50 percent to 90 percent
of a network's traffic, and BitTorrent users can be on the network at any time,
said Comcast spokeswoman Sena Fitzmaurice. That means network congestion from
BitTorrent doesn't just happen in the middle of the day, she added.
"P-to-p traffic doesn't necessarily follow normal traffic flows,"
Fitzmaurice said.
The Internet users who participated in the study may not be representative
of Internet users overall, she added. The users who run the Glasnost tests may
be "heavy users of p-to-p," Fitzmaurice said.
Cox issued a statement, saying it engaged in "reasonable network management."
"Cox's network management practices ensure that bandwidth-intensive applications don't negatively impact our customers' Internet service," the statement said. "Cox allows the use of file-sharing and peer-to-peer services for uploads and downloads, and we allow access to all legal content, but we must manage the traffic impact of peer-to-peer services, as most ISPs do for the benefit of the customer."
Comcast issued a statement repeating its earlier position that it "does
not, has not, and will not block any Web sites or online applications,"
including BitTorrent.
"We have acknowledged that we manage peer-to-peer traffic in a limited
manner to minimize network congestion," Comcast's statement continued.
"While we believe our current network management approach was a reasonable
choice, we are now working with a variety of companies including BitTorrent
[to] move to a protocol-agnostic network management technique."
Comcast announced in March that it would work with the company named BitTorrent
to come up with new network management techniques.
The Max Planck Institute's study seems to confirm testimony by U.S. Federal
Communications Commission Chairman Kevin Martin, who told U.S. lawmakers in
April that Comcast's interference with BitTorrent traffic appeared to be widespread.
Comcast's actions, reported by the Associated Press last October, appeared to "block uploads of a significant portion of subscribers," even during times when the network wasn't congested, Martin told a Senate committee. "Based on testimony we've received thus far, this equipment was typically deployed over a wider geographic area or system, and is not even capable of knowing when an individual ... segment of the network is congested."
The study found BitTorrent interference from 11 other Internet services providers,
in addition to Comcast and Cox, with seven of those in the U.S. But there was
not "widespread" BitTorrent blocking at those ISPs, the study said.
The tests looked at 1,224 ISPs worldwide.
