Which Australian ISPs throttle BitTorrent the most?


news New research published this week purports to reveal which Australian ISPs are throttling their customers’ BitTorrent downloads the most — as well as how they compare with their international counterparts. And perhaps surprisingly, the nation’s largest telco Telstra has emerged as the nation’s best performer.

The research was published by Measurement Lab, an organisation founded by a number of organisations in the US which aims to conduct research into networks and help empower the public by providing it with useful information about their broadband connections — helping to sustain “a healthy, innovative Internet”. Its members include Google and several other organisations, as well as academic researchers.

“When an Internet application doesn’t work as expected, how can you tell whether the problem is caused by your broadband connection, the application or something else?” the organisation’s site states. “It can be very difficult for professional network administrators, let alone average Internet users, to answer this sort of question today. Transparency has always been an essential component of the Internet’s success, and Internet users deserve to be well-informed about the performance of their broadband connections.”

In the research published this week, Measurement Lab used a measurement tool known as the Glasnost test to detect blocking or throttling of BitTorrent and other peer to peer file sharing systems throughout the years from 2008 to 2010. The tool allows Internet users to test whether their BitTorrent software is completely blocked, slowed down or running normally.

According to the Australian results of the test, averaged over two years, the consumer division of AAPT was the worst Australian offender over that period, with 17.7 percent of the recorded tests by the company’s customers suffering from some form of BitTorrent throttling. AAPT’s consumer division was bought by iiNet in mid-2010, but the research does not extend far enough yet to know whether the acquisition has had an impact.

The second-worst offender was Optus, where some 15.2 percent of tests were throttled. After that came iiNet (12.8 percent), Primus (8.8 percent) and Internode (7.4 percent). Coming in at last place was Telstra, with just 6.6 percent of tests being throttled.

The tests aren’t conclusive — with a number of major Australian ISPs being left out, including discount operators such as Dodo and TPG. It is widely speculated in Australia’s ISP industry that TPG is able to offer unlimited download plans at its low prices through advanced control of its network — with customers reporting varying performance at different times.

The Australian results show that in general, the nation’s ISPs throttle BitTorrent downloads less than ISPs in several other countries. In the US, according to BitTorrent news site TorrentFreak, the worst offender was Clearwire, with only 17 percent, but Canada’s Rogers throttled 78 percent of connections, and Tiscali and BT Group 27 percent in the UK. Poland’s UPC throttled a mammoth 97 percent of all BitTorrent traffic, and in Germany Kabel Deutschland achieved a rating of 36 percent.

This area of measurement is fraught with problems, and I have no doubt that at least some of the ISPs would dispute the data which has been collected, as well as the methodology used. With this in mind, we’ve invited iiNet to comment on whether or not it actually does shape BitTorrent traffic. Any comment here may give us some view into how realistic these tests actually are.

Looking at the actual results, it is very surprising to see Telstra coming up as the Australian ISP which shapes traffic the least. We’re not surprised by most of the other ISPs’ positions in the ranking, but we would have expected Telstra to do more BitTorrent throttling than this. On the other hand, it is possible that the telco is shaping less because its broadband plans are more expensive and it likely enjoys more profit margin on them as well (as the predominant Australian wholesaler). Perhaps the big T is trying to maintain its reputation for stellar network reliability? ;)

However you slice the results, however, one thing is clear. All of the Australian ISPs measured showed some level of BitTorrent throttling across their networks. And yet, as someone who browses their pricing plans on their sites continuously in my work as a journalist, I’ve rarely come across anything obvious in their terms and conditions which states that they do so.

To my mind, this is yet another example where Australia’s ISPs need to be a bit more transparent about the service they’re providing. To what extent are we actually getting what we’re paying for, when we sign up for a broadband plan? It’s never really been clear … and the results from these tests again highlight that fact.

Image credit: Lars Sundstrom, royalty free


  1. that’s quite a claim in the current climate and it may be prejudicial for iiNet to answer because, of course, any throttling by iiNet of an Internet application that is predominatly used for copyright infringment will go to whether they have taken “reasonable steps” to prevent copyright infringement for the purposes of the jduciial proceedings they’re currently involved in.

    In fact, if what you’ve reported on is true, it should be considered not only by lawyer’s representing copyright owners but probably the High Court.

    • I hate how they use that.. so why aren’t the ppl that make guns getting sued because of the ppl that use them to kill ppl..

    • Simple answer: “No we don’t throttle protocols used to update World of Warcraft”

  2. For those interested, here is the Glasnost test referred to:


    If you are technically minded, you can read their paper on how their test works:


    (I have had a quick glance; I am not sold on the method because from my (admittedly very quick) reading, it doesn’t seem to go into detail about whether they verified with the ISPs they were testing with were actually filtering or not. I will have a closer read later.)

    I just ran the test; it reckons I am not currently throttled (I am on some fibre link at the moment that I would not expect to be throttled at all). I would be interested to see what results other people get on their home DSL connections.

    I cannot find the source at the moment, but I am pretty sure BigPond have publicly stated they do not throttle BitTorrent at all. I have never ever seen any evidence of it whatsoever. I have used BitTorrent on BigPond connections – and have run it myself on BigPond servers for my own internal usage – and have never had anything that even remotely resembles BitTorrent-specific filtering or throttling.

    I would, frankly, be flabbergasted if they were throttling at all, simply because it seems like a complete waste of time and effort given that they have a pretty good network, they have relatively low quotas anyway, and they meter uploads.

    That’s not to say it has never happened – or indeed, that it is not happening now. But BitTorrent is a complex piece of software. It’s easy to assume that when you’re getting slow downloads it is because of throttling or some other ISP-related problem. It is important to remember that there are many, many, many variables in your average BitTorrent connection.

    By far the most likely explanation for poor performance is just seeds that have terrible upload speed and/or you don’t have a good connection to.

    • By far the most likely explanation for poor performance is just seeds that have terrible upload speed and/or you don’t have a good connection to.

      This right here, the ISP may not be throttling at all, but unstream may be causing slows. Bittorrent is a complicated protocol, you can’t say it’s throttling simply because it’s falling below a certain benchmark.

      • Oh yeh, I forgot to add that one – it’s probably just as likely as my explanation. People flooding their upstream which in turn kills their downstream.

  3. When the group says that an ISP is throttling bittorrent, are we to assume that means it’s definitely not a case of peak congestion, lack of backhaul to a sub-exchange, or QoS during busy periods?

    • Who cares if they looked after there crap we wouldn’t have these problems of slow down… so either way there throttling

  4. its not totally related, but AAPT is/was notorious with WOW players for throttling peer to peer. AAPT constantly denied doing it but come 8pm every aapt consumer i know of that played WOW had horrendous latency and download speeds to torrent sites. seriously unplayable after 8pm when your ping to US servers goes from 350 or so to anywhere up to 28k in a matter of seconds of it turning 8pm.

  5. Something isn’t right here. I’m an iiNet user, and I frequently saturate my connection speeds torrenting.

    • One thing to remember here is ‘throttling’ will usually only apply to single connections, what you should be taking a look at is speeds to each individual peer, and especially their location (maybe it only applies to international peers?)

      Total speed does not mean much given the nature of bittorrent (download from everyone)

  6. < 10% in this kind of test across a large sample isn't a smoking gun. In fact you would expect that there would be a margin of error.

    Anyone running this test over a service that also carries VoIP, is likely to see QoS effects which may or may not present in the results. Lots of people do that. Exchange congestion? Packet loss?

    The application is also Java based; so how do we know the code is sufficiently efficient enough to not be an additional overhead? We don't.

    Lets look at the numbers (which are statistically speaking, trivial) for iiNet:

    Period: 2010Q1 (so a 3 month period)
    Valid tests: 150
    DPI: 13%

    13% of 150 samples is ~22 tests showing some kind of de-prioritisation over the space of THREE MONTHS.

    Last figures for iiNet customer base size? ~652,000.

    This is a non-story Rene! It sounds like someone drumming up business for a technology sale, rather than anything damning.

  7. Oh, I’m sorry, you’re new here, it’s ok. All major ISPs in Australia “shape bit torrent traffic” to some extent. All. Of. Them. Do it.

    That individuals or end-users may get full line speed P2P at anytime is irrelevant & no proof at all.

  8. I’m not really sure how to answer this, what I will say is a year or so ago when I was with iiNet I saw something that ‘looked’ like bittorrent throttling, it also effected FTP, Ventrilo and a host of other niche programs – but I think it may have been totally unintentional, as even the guys in the NOC were totally stumped by my reports (and were able to reproduce my symptons inside the NOC) – they were not secretive with it at all, or even dismissing – they listened to my problems and frankly said “we don’t know” – with that said, reports of specific protocols being slowed on iiNet have all but dissapeared nowdays.

    If iiNet customers want to check if it still happens, try and find a off-shore mirror or something like Ubuntu, find a server that serves both FTP and HTTP, start both downloads at once and see what happens – what should happen is both download at roughly the same speed, but in my case, what happened sometimes in peak periods the FTP would be hideously slow and the HTTP would be fast.

    I actually still have some screenshots from then, example: http://i43.tinypic.com/334nr0l.png – you can clearly see something wrong as only Australian peers give decent speeds (iHug went via peering)

    But again, this probably does not exist anymore. iiNet have more or less completely changed their network since, it’s rather impressive actually, brand new routers too (100Gbps capable I hear!) – the main reason I mention this is maybe the testing dates back to then?

    I’m with TPG, I have not noticed any throttling of bittorrent at all, any times I have noticed ‘throttling’ it has been throttling of absolutely everything at a equal rate because something has stuffed up – if TPG have QOS on their network, it certainly isn’t very agressive, I might even use the word entirely useless ;)

    I ran that test and got ‘no indication that your ISP throttles bittorrent’

  9. I’m curious as to why there are no questions to ISPs named in the article asking them if they throttle and to explain the results.

    Internode Reps are on the record that they don’t throttle P2P:
    Does internode shape torrents / p2p traffic different to http traffic on their SOHO plans?
    > No.

    I am currently shaped on a ADSL2+ plan and can down load from australian sites at 14-15KB/s however my torrents do not go over 6KB/s which leads me to believe that there is some extra shaping on torrents is this the case??
    > No. But under shaping, if you don’t severely back off the bandwidth limits in your P2P client you’ll just cause thousands of TCP threads to choke on each others’ wheaties. i.e. the overheads become the dominant factor.

    Aside from the fact that we DON’T shape P2P (or any other protocols). I’d suggest that because you’re seeing everything stop it’s something to do with your CPE. Maybe your bittorrent client is opening up too many ports and causing the CPE to be unable to open new connections?

  10. There is people who are struggling to survive with no food, water that you wouldn’t even use to clean your feet and have no money and people are worried about this kind of tripe?. People don’t realize how lucky they are….

  11. I would like details on TPG’s throttling. Some torrents are fine others depending on the host are dog slow..

    • Torrents don’t download from different hosts, they only download from users.

      So if you download say a torrent from Blizzard for WoW patches, you are download a little from blizzard (their seeds) but mostly from everyone else in the world who plays WoW.

      If you download a torrent from PirateBay, you download *nothing* (except the torrent file ~20kb) from PirateBay, and everything from “everyone else in the world who downloaded that torrent file earlier”

      In general, the 2 “everyone else in the worlds” should be statistically similar. And theoretically you can’t block one without blocking the other.

      Unless TPG are doing massive scale deep packet inspection:
      – Monitor specific torrent file download locations (say piratebay)
      – Read every torrent tracker from a torrent file
      – Monitor the peer lists coming from a torrent tracker highlited by step 2.
      – Armed with the peer lists, block (or slow, or whatever) connections to just those peers
      That is a lot of levels of remembering what a user is doing between each request to really work at all.

      As near as I can tell, *ANY* torrent throttling, should throttle all torrents equally. If someone knows better, I would really like to hear it (I am genuinely interested in how you would go about selectively blocking torrents from certain sites but not others).

  12. trog:
    “By far the most likely explanation for poor performance is just seeds that have terrible upload speed and/or you don’t have a good connection to.”

    It doesn’t work that way. Their server acts as the seed and only client.
    It isn’t really p2p, it just puts data packet header on the traffic to make it LOOK like p2p.

    It then compares the data rate of packets appearing to be p2p to data packets appearing to be normal packets. If ONLY the p2p packets are slowed, they are being throttled.

  13. PeterA:
    “Unless TPG are doing massive scale deep packet inspection”
    Reading the headers is trivial, it’s how QoS works for ANY packet type. Every router can do it and prioritise traffic as it wants.

    That bittorrent shaping is widespread is evidenced by the fact that all the latest clients off the option of disguising their headers to avoid it.

  14. There are bigger issues out there today.
    World hunger.
    Inflation in the football transfer market.
    I have a sore throat.
    And yes, ALL providers shape bittorrent.

Comments are closed.