Hello,
I'm looking for information on what policies the slurm resume program
calculates new nodes. As I understand, limits can be set like ResumeRate.
For example, how many workers are started without a ResumeRate limit
with 50 short pending jobs? How many jobs can be processed on a node at
the same time? Which job parameters are used for selecting the
appropriate nodes and is the boot-time somewhere taking into account?
So far I haven't been able to find much on the subject of slurm power
saving, maybe someone knows of interesting papers or other sources of
information in this area?
Maybe also papers where it was tested in comparison to another scheduler
scaling technology.
The sentence at the bottom:
https://slurm.schedmd.com/power_save.html
"Booting Different Images
If you want ResumeProgram to boot various images according to job
specifications ..."
But, the power saving plugin can already selects different nodes after
the configuration file with predefined nodes?
https://slurm.schedmd.com/elastic_computing.html
Many thanks in advance.
Regards,
Mike
