> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Matt Doidge
>
> Well I kind of spoke too soon yesterday - after filling the queues just
> long enough for me to get my hopes up maui then started playing silly
> buggers again. It's behaving better, keeping things 80%-90% full (with
> peaks and troughs in free job slot utilisation) , but that's still too
> much waste.
>
Are you sure that's not just it's correct behaviour? If you're
going to run multi-core jobs mixed in with single core jobs
rather than on dedicated nodes, then it's going to have to hold
some CPUs idle until it's got a large enough set to run the big
job. E.g. if you've 8 core node, seven things running on it, and
you've got a 4-core job and some 1-core jobs, but you want to
run the 4-core job, you're going to have to leave that one core
empty until you've got another three. If you keep putting single
core jobs on them all the time you'll never have enough free at
one time to run the parallel jobs.
Ewan
|