Auto-scale amount of workers based on CPU or GPU

[Florian]

Hi all,

I was wondering if there’s a way to automatically scale the amount of worker instances for a render node, based on it’s GPU or CPU current usage?

For example with After Effects jobs, GPU usage can change depending on plugins and effects used.
With a simple project, a render node could have 10 worker instances before hitting the 100% GPU usage. But with a complex project, only 3 instances can be running before hitting the 100% mark, and eventually just stop rendering. Would it be possible to automatically let Deadline balance the number of these worker instances, for maximum performances?

Thanks!
Flo

[anthonygelatka]

most GPU engines will try and use 100% GPU, redshift also uses all available VRAM.

If you’re mixing jobs this may cause issues also, so possibly using deadline command to change the concurrent tasks may work better?

sounds likely you’ll hit issues changing it on the fly as there’s other considerations such as RAM and CPU usage.

This has always been the ideal though, an adaptive and efficient optimization of resource!

Did you submit a ticket to support to see if they have any suggestions?

Ant