Must Reads

CBC’s Marketplace on ISP Speed Claims

CBC's Marketplace features an eye-opening investigation into the misleading ISP claims about consumer broadband services.

8 Comments

  1. I just watched this segment and while I agree that the “marketing” of these services are misleading. However I found the methodology of testing that Marketplace used very unscientific if you are going to make a blanket claim that Provider “A” is faster than Provider “B”.

  2. I agree with Paul. I’ve got nothing to gain by defending Bell, but I’ve had 10 years of dependable, if overpriced internet service.
    Their attempt at testing out download speeds was like polling four people and determining that the results were concrete. It was comedic, but unfortunately laughable.
    The “up to” policy of ISPs on download speeds was interesting, but I found the rest of the story marred by the above-noted feature.

  3. The first thing I will say is this: I would love to see the ISPs put in their place for overselling their bandwidth, not delivering on their marketing claims and arbitrarily shaping packets over customers’ connections. Bearing that in mind, I will also say that this is, without doubt, the most unscientific, inaccurate and sensationalised test I have ever come across. Let me run through and critique the procedures used in this experiment, based on two categories…

    Accuracies and correct procedures:

    1. The telecom consultant accurately notes that cable connections are community-based and that their bandwidth often depends on how much collective bandwidth is being used on the local node for your neighborhood.

    2. He also states that distance from a DSL exchange point will affect your bandwidth. However, this should not ever be a problem as DSL ISPs should not be selling connections to nodes that are too distant from the exchange building to achieve reasonable and claimed bandwidth rates. I will give them credit for this, as Bell shouldn’t have been selling service as distant as this in the first place.

    3. They correctly identify that ISPs are limiting their customers’ bandwidth rates because they are overselling their bandwidth.

    Inaccuracies and incorrect procedures:

    1. They download the Ubuntu Feisty Fawn release (approximately 700MB) via HTTP. This, in itself, renders this test useless and meaningless because they have no clue what sort of load the server is currently undergoing or what the upstream bandwidth of the said server is (which could very well be slower than the downstream bandwidth the customers are paying for and actually getting). A single network connection is only as fast as its slowest point (also known as a “bottleneck”) and this fact seems to evade them. While this may give a very rounded ballpark figure, it is nowhere near scientific accuracy.

    2. They download the iTunes software application and assume it’s an accurate test. Who knows what sort of load Apple’s web servers are experiencing during these downloads?

    3. They’re relying on the accuracy of extrapolation-based software time and speed estimation (and that of Internet Explorer’s no less, which is arguably the least accurate of all), which has absolutely nothing to do with the actual bandwidth rates they’re experiencing, because it’s EXTRAPOLATING.

    3. At no time, do they account for any typical network overhead in any of the tests.

    4. They infer that the laptops are not the bottleneck because they’re all using “new computers” (which we assume aren’t defective). Thanks, I really needed you to tell me that, since they are all at least 100 Mbps Ethernet connections (or possibly even 1 Gbps) and the fastest of the Internet connections maxes out at 7 Mbps.

    5. An interviewee, Helga, claims that she can “go make coffee” when she’s “downloading stuff”. Well, there you have it folks, if someone can “make coffee” while “downloading stuff”, then you know for sure that she’s being totally ripped off! Can it get worse then this? Read on…

    6. The telecom consultant claims that the wiring and cabling, modem, router or network card could be causing a bottleneck. With the exception of the modem, that’s pretty doubtful unless the device is defective or seriously misconfigured.

    7. They claim that “Joe” is getting only 10% of the bandwidth his ISP claims to provide, a conclusion arrived upon by…well, we’re not really sure, because they don’t tell us (not that it would be an accurate measure if they did).

    8. They seem to think that the fluctuations in bandwidth rates are due to their ISP (which is possible, but unlikely most of the time) rather than the nature of networking (overhead, server loads, community node connection load, bottlenecks at exchange buildings, technical outages and what-have-you)

    9. After configuring a 7Mbps Bell connection at a next-door residence to someone who claims to have problems with his DSL problems (which turns out to be due to being too distant from the exchange building), they download a completely different file (who-knows-what from who-knows-where) and conclude that the bandwidth has the same shortcomings due to distance. No kidding.

    10. Not once do they ever account for any services or background applications consuming network resources on the PC.

    11. There’s no mention anywhere of what sort of upstream bandwidth the customers are being sold. You’d be surprised at how substantially your upstream bandwidth affects your downstream bandwidth.

    Unfortunately, most of the conclusions they draw are not because of the big, bad corporate bully yanking customers around, but rather the vast amount of misunderstanding surrounding Internet infrastructure, normal network overhead, locations of bottlenecks and other factors that all contribute to bandwidth at any given time.

    Here’s how they could (and should) have made this a scientific and accurate test…

    1. Configure a server with dedicated, guaranteed upstream connection equivalent to all of the customer connections’ downstream bandwidth rates combined (or faster) and run downloads of copies of the same file simultaneously OR configure that server with an upstream connection that is equivalent to (or faster than) the highest common denominator of the customers’ downstream bandwidths and run the tests separately (but of course, they couldn’t do that, because it wouldn’t make for the same sensationalised and entertaining “race”).

    2. Use precisely the same equipment everywhere: same PC models, same routers, same network cables, same modems for the same types of connections and configure them all the same way.

    3. Run more tests than just a single HTTP download and run them multiple times. Run ping tests with names and IPs separately, DNS lookups using the ISPs DNS servers and also other DNS servers, run download AND upload tests over multiple protocols, multiple times (HTTP, FTP, SCP/SSH, VNC transfers, VPN transfers, proprietary application protocols – anything else you can think up).

    4. For crying out loud, use PROPER network analysis tools like Wireshark/Ethereal instead of browser-based extrapolated bandwidth and time estimations.

    CBC did an extreme disservice to its viewers by offering sensationalization instead of accuracy. If ever the ISPs come under fire for their wrongdoings (which they should), poking missile-sized holes into this “investigation” will be an excellent weapon for their defensive arsenal.

    Sorry, Michael, I expect a little more from you than to say that this is “eye opening”. The only “eye opening” thing in this investigation is how incredibly innaccurate it is and how they couldn’t be bothered to be the least bit scientific about it. To be fair, I’m sure you’re a busy person and don’t always have time to delve so deeply into auxiliary projects such as your blog. Nonetheless, I really appreciate you bringing things like this to your readers’ attention.

  4. Sympatico facts
    Eye opening, DslReports good bad and ugly report:
    [ link ]
    Notice Sympatico is the worst in Canada

    Eye openng, Sympatico user reviews:
    Six Month Rating: 59%. The LOWEST of all Canadian ISP\’s.
    [ link ]

    You are correct, Marketplace could have done better to expose the truth in sympatico\’s misleading ads that played, lies from the techs and management, and the piss poor service it gives people.

  5. X-TELUS
    I understood from the MarketPlace video, that all the users downloaded the same file at the same time.

    What doesn\’t matter then, is whether the servers are \’under a load\’ because the same file on the same server will produce the same performance environment for all four of the downloads. Exception being for files from a Server-Farm.

    For people with consistent bandwidth performance issues I understand that a \”Signal Booster\” could improve the connection — Search on MODEL – 48409500100 VENDOR -General Instruments.

    Agreed that the \’test\’ is not at all scientific, and is a simplification of \”Internet Speed.\”

    The \’Internet/Computer/Software\’ industry is rife with \’bait & switch.\’

    I\’d like for CBC MarketPlace to do a much more comprehensive documentary, something such as Bill Moyer\’s PBS \”The Internet At Risk.\”

    X-TELUS

  6. Unit problems?
    Just a thought, doesn’t Mbps mean mega BITS per second while most file downloads measure in BYTES, which means that the download speed (well, the number) would be 1/8 of the advertised bandwidth??

  7. End User
    After many years of putting up with Dial up in Northern Ontario, it was a great pleasure to have our Government splash out 1 million dollars to install wireless internet in our area. Finally high speed internet. WHAT A JOKE!

    After one year of fighting so called technical bimbos to get the equipment up to date and working; I now find myself fighting the ISP because they have overloaded the entire system.

    A system that promises (UP TO) 1.5 Mps downloads, ya if I am the only one on the system and I am using the server in my own home. This is just a complete and utter failure on the behalf of the CRT.

    Internet should be sold and paid for based on speed and
    bandwith measured to the end user. If I get dial up speed while using it, I pay for dial up. If I get 7 Mps, I pay for it. But NO! I am a Canadian Cow, who pays $54 per month for (UP TO) 1.5 Mps that has proven to be more like 300 Kps at best when God is with me.

    There isn’t a lawyer in Canada with big enough Coconuts to take on these thieves and the CRT is…well you know, no comment.

    I am always pleased to visit the mega cities, breath in their polluted air and risk getting mugged or raped to enjoy there access to true high speed internet.

    Will it ever change? No. Not until all of us, collectively choose to stand up in court and say, okay time for a refund. I await the day.

    Until then, I must put up with my just better than dial up service and hope that my Government does not continue to burn my tax dollars on another mega waste project of providing so called internet service.

  8. Yes it would be nice for all the tests to be done using the same equipment connecting to the same server. That doesn’t reflect the real world, I currently have telus and reasonable often get the speed I pay for (6 Mbps or about 700 KBps). I am changing providers because telus blocks ports limiting what I can do with the service and there customer service has annoyed me for the last time