Post #3057402
2026-05-21 14:44 UTC
Replies (3)
-
@penguin42@mastodon.org.uk 2026-05-21 14:45
@globalc@chaos.social IMHO the best solution is a differential case - measure the power without the process, then measure with it.
-
@globalc@chaos.social 2026-05-21 21:04
@penguin42@mastodon.org.uk That's good input - trying with another cpu-bound workload, i.e. stress-ng or a pi-calculator. I'm also heavily using virt-manager, but did not see the needed piece for the usb - it's not really device trough but QEMU acting as host, and serial ontop. True also on the "just switch single applications on/off, compare overall consumption". But can't do on realworld systems providing a service. Also not interacting apache-php processes on same system.
-
@globalc@chaos.social 2026-05-22 08:39
@penguin42@mastodon.org.uk That was it apparently - providing /dev/urandom by itself making more load for the VM. Switching to another load, a pi-calculator, I see the attribution as expected. I'll update the article, and to the least also give that our KVM folks to consume and ask if they are aware. :) Thanks for the idea!