Data collected by the BitTorrent client Azureus shows that Comcast might only be the tip of the iceberg when it comes to BitTorrent throttling ISPs. Early findings show that customers from quite a few other Internet service providers experience an unusually high amount of TCP-resets.
ISPs have been throttling BitTorrent traffic for quite a while, but only since the Comcast debacle has this been picked up by mainstream media.
A few months ago Azureus petitioned the FCC, which led to a FCC hearing in February. One of the complaints from the commission was that there is little data available on the scope of BitTorrent throttling, a gap Azureus now tries to fill by collecting data on the prevalence of TCP-resets among ISPs worldwide.
Last month Azureus published a plugin through which users can help distinguishing the good from the bad ISPs, and today we have a preview of some early findings. A massive 1,000,000 hours of data from over 8000 users has been collected over the past few weeks. The preliminary results again confirm that Comcast continues to use TCP-resets to manage BitTorrent traffic on their network, but they are not alone.
The rest of the Vuze/Azureus report (pdf) includes the median reset rates for hundreds of other ISPs
|Brasil Telecom Santa Catarina,||Brazil||17.43%|
TCP resets seem to be more common for American ISPs, and Comcast leads the bunch. The Azureus team has sent a letter to Cablevision, Cogeco, BellSouth and AOL, where they request that the companies are open about their BitTorrent throttling practices. Thus far, the ISPs have not responded to the letters.
At the bottom of the list we see the good ISPs, mostly from Europe. There are other ways to throttle BitTorrent traffic, besides using TCP-resets, a list of ISPs who are known to limit BitTorrent traffic is available on the Azureus Wiki.
|Telecom Italia France||France||2.53%|
|Orange Nederland||The Netherlands||2.57%|
It has to be noted that the data gathering techniques Vuze uses are far from optimal. The plugin detects all TCP resets on a connection and doesn’t make a distinction between BitTorrent and other traffic, and there is no control group.
The Azureus/Vuze team will continue to collect data, and stated:
“We believe that there is sufficient data to suggest that network management practices that ‘throttle’ Internet traffic are widespread. At a minimum, more investigation is required to determine whether these resets are happening in the ordinary course of business or whether they represent the kind of throttling practices which target specific applications and/or protocols, harming the consumer experience and stifling innovation.”
The preliminary results presented here do indeed indicate that Comcast is not the only ISP that uses TCP resets to slow down BitTorrent traffic. People are encouraged to continue using the plugin so more robust data can be presented in the near future.