AWS Thinkbox Discussion Forums

Set core limit used per task when job is submitted?

When I submit a job I see the “concurrent tasks” option, if I set this to 4 concurrent tasks when I submit a job and I have a 16 core machine in the farm, each task should grab 4 cores to render and if I have a 4 core machine each task will grab 1 core to render.

But lets say I have a task which its optimal amount of cores is 2 cores per task. Is it possible to set up a max core limit per task so my 16 core machine will grab 8 tasks and my 4 core machine will grab 2 tasks? This way I will keep the optimal amount of cores per task constant throughout my farm.

Many thanks!

This isn’t quite accurate… Each concurrent task will spawn its own render process. There is a clamp to let you limit that count on a per-Slave basis:

This does not impact thread counts that you set at submission time for those applications that support it.

Now, there is also CPU affinity which only clamps the cores the processes are going to run on. If you submit a job with 16 concurrent tasks but the Slave is only allowed to use 1 core, they’re going to use that one single core.

Any thread/core affinity is going to be done at the Slave level and the render plugins don’t have access to that. The GPU affinity is significantly different though as we must control that at the render plugin level.

Privacy | Site terms | Cookie preferences