2012/8/9 Ole Tange
<tange@gnu.org>
I just swapped a machine to death by starting 1 jobs per CPU on a 48
core machine. The problem was that each job took more than 1/48th of
the memory.
Another approach could be to avoid starting jobs when the system is swapping.
It take a very little resource to get the current number of pages swapped in or out:
$ grep ^psw /proc/vmstat
pswpin 83218
pswpout 179322
So after each job is started look if swap in/out has changed. If does not change start a job more.
I don't know if this tactic will work but it worth a try. On other OS you might have to use sysctl or something.
/hans