Results of latency test, list of computers tested for use with LinuxCNC

Would an output like this be helpul? greez
chris

BeagleBrainz wrote: It would be interesting to see the fstab for the HDD vs the SSD. One would think the SSD would have a lower impact on latency.
Any details of the model of each ?

What is the amount of cache for each device ?

Bearing in mind if I have this right:
What are the diffs if you start the actual latency test and then start open GLX gears & firefox. As I see it open GLX & Firefox after would show any issues with file access.
Then we could see the affect of having noatime,nodiratime in fstab and not having those options.

I could add an option to show fstab and lsblock info to the script. There could also be an option to run the script for an amount of time w/o any glxgears or firefox. Would that be helpfull?

greez
chris

Usually when I run a latency test I start with no programs running, start the latency test, then start a few instances of glxgears, open Firefox and try a few other things.
This should test what affect opening and starting a few programs has on latency.
Currently, if I have it right, the only IO access in the script is writing to the file stores the results. There’s no opening and closing files and no starting the running and closing of any programs ?
I don’t know if that would make much difference.

You are correct: Other than glxgears, firefox and writing .lat and .hw file.
I guess that resembles a lot of how axis will work ?

If needed, I can add an file io load. But as laid out above: I do not believe that is representing how LinuxCNC works.

Greez
chris

The idea of the latency test is to stress the machine, whilst running the test. I think it's actually mentioned in the docs or somewhere, to open & close windows & move files around.

You can even influence the results by changing the IO scheduler. Mount options can have an affect as well.

I'll check the possibilities to do a suitable io load. Will put the discussion over in this thread .