Tool reveals ISP traffic-shaping


ShaperProbe is a research tool from a team at Georgia Tech that detects and characterizes "traffic shaping" by ISPs, determining which US ISPs deliberately slow down their customers' traffic, and in what way they do so. They've published a paper on their findings, which indicate that cable ISPs do much more "shaping" than telcos, with Comcast and Cox being the most aggressive, even slowing down the traffic you send, in addition to the traffic you receive. Georgia Tech researchers recently demonstrated another tool, Kermit, which also measures ISP performance. Both projects were funded by grants from Google, who have a clear interest in helping Internet users determine which ISPs practice throttling and shaping (notwithstanding Google's Net Neutrality sellout with Verizon).
The use of shapers can also be dictated by the underlaying access tech. The researchers note that DSL providers can "dynamically change link capacity instead of shaping, while a cable provider is more likely to shape since DOCSIS provides fixed access capabilities." It can also be applied to the entire Internet link or more narrowly to specific kinds of traffic-- which, as the researchers note, is "relevant to the 'heated' network neutrality debate."

The research shows that Comcast, Road Runner (from Time Warner Cable), and Cox all use downstream shaping--but only Comcast and Cox also use upstream shaping. Neither AT&T nor Verizon shape in either direction.

New "ShaperProbe" tool detects ISP traffic shaping

End-to-end Detection of ISP Traffic Shaping using Active and Passive Methods

20

  1. So my choices are fast but “shaped” (RoadRunner), fast but evil (AT&T), or slow but not shaped or particularly evil. Sigh. I’ll stick with RoadRunner for now, I guess.

  2. This is exactly why I LOVE my Monkeybrains ISP.

    If you live in San francisco, you have to do go with Monkeybrains.
    They are a local independent ISP and they dont do any shaping… and so fast.

  3. Are you referring to Verizon/MCI with “slow but not shaped or particularly evil”? Verizon? FIOS is Verizon….fastest internet I ever had.

  4. So can we get the mainstream media to butcher this for more attention? “Is your internet provider stealing the access you pay for? Tune in at 11:00.” I’d love to see major providers stop this practice.

    1. I’d love to have some choices other than “major providers”. They all suck.

      I wish I could get something like Monkeybrains in the East Bay…

      1. Yes, they do all suck. Everywhere. It irritates me to no end that monopolies were lobbied for and granted, but never regulated in any meaningful way. Our cable/internet bill has gone up $12/month in the last year. For what? An even slower connection than it was last year.

        That’s the story in America these days, isn’t it? Pay more, get less, and you damned well better feel grateful to have any service at all.

  5. This might explain why I could never get a good torrent connection going on Comcast to save my life.

    1. The odd thing is my Road Runner (which is supposedly “shaped”) is very fast at torrenting. I can get over 1,000KB/s (8mbps) at a good torrent – which is enough to cause my router to go haywire. So they must not be shaping it too aggressively.

  6. I’ve had Cox cable internet for about five or so years. Personally I’ve not had a problem with them, could they be slowing my connection… apparently. But I haven’t had any of the nightmares with them I’ve read from people with Comcast. Though it’s not like there’s too much of a choice in ISPs where I live it’s either Cox or Qwest DSL. Cox actually raised speeds a bit without raising the price (at least for now)

  7. Its important to remember that Comcast advertises this shaping as a feature: http://networkmanagement.comcast.net/NM-update-03102011.htm

    Their “PowerBoost” service allows a short period of transfers above the user’s base speed. Many other cable companies use similar techniques.

    The actual technique is a “leaky token bucket” shaper: a user gets a sustained rate, and can exhaust the bucket for short above-the-rate transfers. Its built into many of the DOCSIS head-end units thus it is natural for cable companies to use it.

    EG, my home connection will transfer the first 20 MB down and 10 MB up at a higher rate before dropping down to the rated bandwidth. The result is that for web surfing and the like that things seem much faster than if limited to the rated bandwidth, without requiring that the company provide that above-rated bandwidth for big file transfers.

    The disadvantage is that it may confuse things like Netflix etc, if it measures the initial download during the burst phase and then when things don’t keep up, it has an awkward transition to the sustained rate.

  8. I’m a Comcast customer. When I start a new torrent, my non-torrent bandwidth drops to zero. I cannot torrent and browse the web at the same time. The torrent will in fact download/upload normally; Comcast is not denying the packets. But the “punishment” for using BitTorrent is loss of web access. This doesn’t happen when downloading Steam content or other large file transfers. Only torrents. As a consequence, if I really have to torrent something, I do it at night while I’m sleeping.

    1. It sounds like you need to adjust your settings with whatever tracker you are using for torrents. I use transmission and it lets you adjust the bandwidth allocated to the torrent and what is left over for normal internet browsing.

      You need to read the FAQ on optimized settings.

      1. Yep, it is a classical oops move to saturate ones outbound bandwidth (much lower then ones inbound unless one is on a fiber service of some kind). This then results in all kinds of timeouts when trying to browse sites and such as the request packages can not get out because of all the torrent traffic.

  9. Could this differ from one city/state to another? Little to no evidence of shaping thru Comcast.

  10. MrWooster: Your problem is not Comcast, your problem is BitTorrent…

    If you configure it wrong, BitTorrent will effectively consume all available bandwidth. With how TCP works, if you have 4 flows that all want to consume the available bandwidth, an additional 5th flow will not get much, and your web surfing etc would take time to even get up to speed.

    This is especially bad when combined with Buffer Bloat, so BitTorrent’s TCP flows will fill up the uplink buffer to its capacity and you will now see 1s+ latency for everything until you quit BitTorrent…. Which, to put it simply, sucks.

    You need to limit the BitTorrent transfer rate to less than your available bandwidth and no more than 4 TCP streams, and (if your client supports it), turn on the UDP transport (which is friendlier than TCP and designed not to cause problems with buffers).

  11. clearwire Is A good alternative if you live in the US. Take your internet with you wherever you go at decent speeds.

  12. Clear in Las Vegas is AWFUL! I get throttled for any slight usage. Wireless simply does not have the bandwidth for many users that try to use their provisioned speeds. Currently 0.58Mbps down/ 0.71Mbps up, via speedtest.net.

Comments are closed.