Monday, May 30, 2011

Who Shapes Traffic in the US?

New technology, developed by Partha Kanuparthy (picture) and Constantine Dovrolis from Georgia Institute of Technology, allows to detect if ISPs are using traffic shaping,

See "End to end Detection of ISP Traffic Shaping using Active and Passive Methods" - here.
"First, we develop an active end-to-end detection mechanism, referred to as ShaperProbe, that can infer whether a particular path is subject to traffic shaping, and in that case, estimate the shaper characteristics. Second, we analyze results from a large-scale deployment of ShaperProbe on M-Lab over the last few months, detecting traffic shaping in several major ISPs. Our deployment has received more than one million users so far from 5,700 ISPs. Third, we modify the Shaper-Probe detection algorithm so that it can be applied passively on the traffic of any TCP-based application".
So which ISPs are using traffic shaping? 4 case studies are presented in the paper; see also the table below:
  • Comcast - "We observed many shaping configurations in our observations between October 2009 and May 2011" (no need to repeat the Sandvine-Comcast history)
  • Road Runner - "94% of upstream runs did not detect shaping, while 64% of downstream runs found shaping"
  • Cox - "we found Cox to have a significant number of upstream shaping detections of 20%" (see - "Procera: Cox Generated $2.2M in 2010" - here)
  • AT&T - "We look at properties of the 10% of AT&T runs which were diagnosed as shaping.. we see that about a third of these runs show a strong shaping rate mode and an associated burst size mode"

1 comment:

  1. Its unfortunate that the study authors don't really have a grasp on what is and isn't used for traffic shaping. They do correctly identify that Comcast and many other DOCSIS systems use bursting to increase the temporary throughput of users but that has nothing to do with traffic management and in cases where contention is a concern "boosting" or "bursting" of traffic is one of the first things to get turned off. It also has nothing to do with Sandvine, which operates at layer 3 (or higher), and is handled by the CMTS and the modems as part of DOCSIS. All in all a good piece of study that is critically flawed by lack of understanding of the data the authors collected.