Multiple runs (warm up run)

[pkirsche]

Is your feature request related to a problem? Please describe.
Everytime I run a manual speed test, I get around 10% better results than the periodic automatic runs. I'm using always the same server and observed this behaviour also when running the manual test only a few minutes before and after the scheduled test.

Describe the solution you'd like
Maybe we can have an option for multiple speedtest runs? Let's say on every scheduled run, there is not just this single test but another test where the second one then is used for logging and saving.
I would call it an additional "warm up" run :)

[svenvg93]

Hi @pkirsche,
What is the cronjob settings you are using? Other users (see #1263 (comment)) have seen less results when running the on the hour.

The theorie is that there are more users testing against that server on the same time, giving less results. Since your results are better before and after the scheduled tests it might be the same. Can be worth trying the set the schedule on the random moment after the hour mark. For example I run mine 17 minutes past the hour.

[c11umw]

It could possibly also relate to bufferbloat where the device calling the SpeedTest isn't being given the network priority to perform the test in full due to other network traffic already queued up or happening at that the time.

You can test it here: https://www.waveform.com/tools/bufferbloat

In theory it should be fixable with QOS enabled on your router (aka "Smart Queues" on Unifi routers)

[alexjustesen]

#1444 coming in the next version might help this.

[pkirsche]

Thanks for all your replies and sorry for my late reaction.
@svenvg93 Your hint with a different testing-time has improved and stablized the results for me quite a lot. Thanks!

@c11umw Nice test page, didn't know that such an easy way exist to test for buffer bloat. After running the test, I got an "A+" grade, so this seems not to be the problem. But thanks anyway for the interesting website.

@alexjustesen Sounds good, I'm looking forward to the next version. Please keep up the great work and thanks for all your efford until this point! I really like the project, its very useful!

[alexjustesen]

@pkirsche v0.20.x series includes pinging an IP address or domain prior to running a test, has this helped with "warming" a connection?

[pkirsche]

@pkirsche v0.20.x series includes pinging an IP address or domain prior to running a test, has this helped with "warming" a connection?

Looks very good now. 👍 Moving to another time schedule already made the results more constant the last weeks, but with the new release, I added a domain to test and the results are absolutly stable and at the limit of the connection now. Very impressive, it seems it needs some traffic before the messure itself to "wake" my connection up.

Thanks!

[alexjustesen]

Fixed with #1444 and #1493