have a crash, rapid reboot stateless provision, and let another user
get the node, than a job start to crawl and waste time on the system.
this is what bugs me about anti-swap believers: they conflate swap,
or even swap use, with thrashing. then they have to derate the mem
capacity of nodes in order to make up for it (ie, to provide some
slop for non-job-process uses.)
swapout is good. it's how the kernel keeps ram warm, rather than
letting cold pages *waste* your ram. swap IN can be bad (thrashing),
but the main point is to toss cold pages into the attic
that you're not going to use soon or ever.
if you have tons of ram, go for it! normal people want to optimize
the use of ram - potentially even to have prempted processes
in suspended animation.
we also run with overcommit=2.
regards, mark hahn.
_______________________________________________
Beowulf mailing list, [email protected] sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
