Speed Testing Your Connection
Yesterday Minnesota Public Radio did a story on mapping broadband service in Minnesota. Affordable, high-speed Internet connections are becoming a critical component for educational and economic development throughout Minnesota and the world. Both the Blandin Foundation and the Minnesota Ultra High-Speed Broadband Taskforce are working hard to increase the availability of broadband throughout the state.
Because it isn’t cheap to bring broadband service to rural areas, knowing which areas are most in need is necessary to determine where to best spend limited resources. The Minnesota Commerce Department is working with Connect Minnesota to map Internet connection speeds throughout the state and is promoting the Connect Minnesota Speed Test as a way for consumers to check on the truthfulness of their ISPs. While this is a great goal, there are significant technical problems with the Connect Minnesota Speed Test.
The first problem is the coding and assumptions present in all Ookla speed tests. Because the speed tests assume you are on a DSL or cable connection, they automatically assume a huge asymmetry in upload and download speeds. Therefore, they give false results for any type of connection other than DSL and cable connections. Don’t use this to test the connectivity for your servers colocated at an ISP or your office T1 or Metro Ethernet connection. Our engineers have pulled down 230Mbps to their workstations here at the office and Ookla speed tests are unable to calculate that type of speed.
The second problem is that the Connect Minnesota Speed Test site itself is clearly limited by a 10Mbps connection. No Internet connection ever performs at 100% and there is always going to be some routing and other overhead wrapped around the actual data itself. This type of speed test will always underestimate the actual speed of any given connection. This is an old speed test issue. Internet routing hasn’t changed and all the caveats Peter John Harrison wrote about in his 1999 SpeedTest still apply.
Finally, and here is my major problem with the Connect Minnesota Speed Test, it is coming from a provider and machines in Texas! This means that it isn’t really testing the speed of your Minnesota Internet connectivity (even with all the above provisos) it is testing the speed of your connectivity to some provider in Texas. There is a lot of network between here and Texas. The packets for the speed test are traversing different networks and providers on their way to and from the server in Texas. They may be affected by various routing issues on networks that have nothing to do with your ISP and that your ISP has no control over.
If the Minnesota Commerce Department wants a more accurate picture of broadband connectivity within Minnesota, they need to locate the speed test on a well-connected server within Minnesota.
Of course, anyone who really wants an idea of the speed of their connection should not just be using an Ookla speed test. Basic FTP gives you a much better test to determine how fast you can pull down a file from a remote site. In fact, ipHouse has files on our servers specifically for customer speed tests. More accurate tools exist for calculating bandwidth but they aren’t easy for the average consumer to use. They almost always require command line access on two machines on each side of the connection you are testing. If you are an ipHouse customer and want to know the speed of your connection, just contact our support team and we will let you know the test options available for your connection and how to see bandwidth usage throughout our network.
Comments are closed.
about 1 year ago
Tarnation! What a dumb mistake. It’s too bad — the policy type people were really counting on this test for some insight.
about 1 year ago
Aileen,
Here’s another speed test that gets a lot of press:
http://www.speedmatters.org/
It’s maintained by the CWA (Communication Workers of America). Does it have similar issues? It targets users across the US. I had assumed it somehow made allowances for location – but I’m betting it doesn’t. The CWA reports get a ton of press and I’m sure also has an impact on policy decisions.
Thanks for being the smart folks to really look at this!! Ann
about 1 year ago
Ann,
Although I didn’t see a logo, the Speed Matters test appears to be another Ookla test. The FAQ for the Speed Matters test states that, “The speed test has access to a fixed number of servers around the country, automatically choosing the one closest to you to conduct your test. But the server used may be farther from your actual location than is optimal. Other speed tests may have access to far more servers, allowing them to test your speed with a closer one and giving you a faster result.” So yes, it is better than an Ookla test that just uses one server in one location but still may not be accurate.
Unfortunately the only explanation they offer for why downloads speeds are so much faster than upload speeds is that your ISP must be configuring your connection to favor download speeds. Just now my desktop machine tested at 29Mbps downloads but only 1.9Mbps uploads. I know that huge discrepancy is being caused by the test *not* my ISP.
I guess the real question is how are people trying to use these statistics? Are they trying to figure out which areas of the country have just dialup? Which ISPs have oversold their connectivity to their end users or their bandwidth out to the Internet? Are they concerned about bandwidth or connectivity or both? Some national networks have low Mbps prices in part because they have fat pipes that aren’t very well connected across the country. Are they more concerned about home access or business access? (Assuming home users are more likely to be on asymmetrical DSL or Cable connection and businesses are more likely to be on symmetrical T1s and Metro Ethernet connections.)
The issues with the tests are more or less important depending on how the test results are being used.
about 1 year ago
Ann,
The Speedmatters test is another flawed example. In fact, it has two problems; one being there is not a truthful way to correlate the zip code to the test. I put in a zip code of 59928 and got a whopping 30mbps download. Funny, I’m actually in 55401 and not in Polebridge MT. Besides the geographic reliability being in question, the other error I witnessed was the upload being “limited” to 7mbps. As you know, we have synchronous connectivity here. I should have seen 30mbps on the upload side too..
Bil
about 1 year ago
Hmmm. I suspect (OK I know) that the CWA reports that come from the Speedmatters tests find their way into the hands of a lot of decision and policy makers’ hands.
That’s the intention of the Minnesota maps too. It is a tool provided to the Minnesota Ultra High-Speed Task Force to help them make recommendations to the Legislature.
I hope the goal of the test isn’t to highlight un-served areas (dialup versus anything better). That would be a sad state of aiming way too low. I’d like to have the maps help determine which areas are underserved, served and served well – so that we can pull other data and see if we can draw some correlations between tier of service in an area and economic vitality and quality of schools and health care and so on – including ability to work from home and start home-based businesses. So home and business access is important.