I’m trying to resolve what I would consider a common problem. For simplicity, we have two types of renders, I/O bound and cpu bound. I/O bound renders are like krakatoa and fusion comps that load lots of images. And cpu bound renders are the normal raytracing renders like brazil, vray, mentalray.
We have a few machines that are beefy on the cores. But they are mostly idle when rendering I/O bound jobs.
So we could add a few more slaves that can only render those jobs. But we also want a slave that uses all the cores for the raytracing jobs.
How can we have slaves that only turn on when the “main slave” is rendering an i/o bound job.
Am I making sense?
Hey Ben,
This kind of “slave scheduling” isn’t supported by the multi-slave feature. The slaves run independently of each other, so they’re not aware of what the others are doing. That could change in the future.
Ideally for now, you would be able to set up your pools/groups so that the slaves can always be running, and they should be running the types of jobs that wouldn’t interfere much with each other. You could also set up slave scheduling in power management so that the IO bound slaves on the beefy machines only run during the day, leaving the raytracing slave in full control of the machine during evenings.
Cheers,
Ok, so I think we are stuck with longer raytracing renders then. Or the i/o bound renders need to have concurrent tasks.