Confused Bird Forum
Use TestMy for download speed diagnostics - Printable Version

+- Confused Bird Forum (https://confusedbird.com)
+-- Forum: Technical (https://confusedbird.com/forum-1.html)
+--- Forum: Mobile Broadband (https://confusedbird.com/forum-2.html)
+--- Thread: Use TestMy for download speed diagnostics (/thread-9.html)



Use TestMy for download speed diagnostics - Seán - 23/01/2020

Anyone that gets a chance to test a high speed Internet connection will almost certainly head to Ookla's Speedtest.net or use their app to test the speed.  With fixed line services such as VDSL, cable and Fibre, that link (if working properly) is generally capable of delivering what Ookla reports. 

To reduce the chance of non-network related activity such as high CPU usage or the Antivirus buffering and releasing data during analysis affecting the speed, Ookla's methodology has a few tricks.  It filters out a portion of the slowest part of the test as well as any brief spike, then averages the rest.  It also makes lots of simultaneous connections to the test server to get the peak speed.

That methodology may make sense with a Gigabit and fast cable connections to see what they can get up to, like speed testing a car.  However, when it comes to erratic connections such as cellular data and even Wi-Fi, filtering out throughput dips can lead to test results much higher than what the link is capable of sustaining.

To demonstrate this, I purchased the Internet Speed Meter App that shows a real time traffic graph in MB/s.  After running multiple speed tests with the Ookla Speedtest App and on TestMy, these are the results chosen when the link throughput was erratic.  I drew in the blue (download) and red (upload) lines, which correspond to the test result in MB/s.  As TestMy runs upload and download tests separately, this screenshot only contains the download test:

[attachment=9] [attachment=10]

While the left connection throughput clearly got up to the results shown, it certainly did not sustain either the upload or download speed during the test.  Basically, Speedtest shows what the connection throughput gets up to, not measure what the connection can realistically deliver. It's much like how getting up to 50km/h in rush hour traffic does not mean you will drive 50km in an hour.

TestMy uses a very simple methodology.  It times how long a known block size takes to download or upload and calculates the result.  This gives a much more realistic test result on erratic connections, particularly when testing with a large block size.  This is much like timing how long it takes to drive a mile in rush hour traffic instead of seeing what your speedometer gets up to.

With the blue line I drew in on the right representing 22.8Mbps (2.85MB/s), the graph area above the line would fill the gaps below the line.  At this time of testing, this is the actual download speed the connection was capable of sustaining.  Had I ran Speedtest at that time, it probably would have given around 30Mbps based on what the throughout got up to.  

TestMy can also run Multithread tests for testing the capacity, however, I recommend sticking with the regular Linear test, particularly during times of congestion.  For example, if TestMy cannot deliver 5Mbps with a large block size, don’t expect to be able to stream at 5Mbps minimum.

When testing on the Three network, it’s also worth testing for traffic shaping as explained here.