[DGD] Hydra

[bart at wotf.org]

On Sun, 16 Sep 2018 17:45:51 +0200, Felix A. Croes wrote

<snip>
 
> I think what we are seeing here is a system trying to decide
> whether to run everything on one CPU or two CPUs, and doing a lot of
> catch-up copying when changing strategies.  When the number of
> active CPUs is unambiguously higher than the core count of a
> single CPU, it performs worse for 50 recipients, but better for
> 100 recipients.

Linux has different schedulers. If it is useful for you, I could rerun the
test with those different schedulers. I'm not sure which one was active while
I ran the test, as this is normally managed by tuned on this system. This is
because my everyday use of this system is as an esxi server which among other
things runs one fairly big vm which has direct access to some pci-e devices
for storage and networking. I can also boot that vm 'physically', which I did
for running that test, and which I at times do when wanting a linux host on
bare metal and with lots of cpu cores.

If its interesting, I could also run the test in a VM on the same hardware
(but limited to 4 cores and 8gb memory)

Bart.
--
https://www.bartsplace.net/
https://wotf.org/
https://www.flickr.com/photos/mrobjective/