[torquedev] Some dev. notes for the current discussion around ncpus/procs

"Mgr. Šimon Tóth" SimonT at mail.muni.cz
Tue Mar 2 12:34:54 MST 2010


I would just like to note that I have a reasonable amount of code, both
in the server and the scheduler that implements better work with
resources (and scheduling in general).

They cover a lot of stuff from simple things like generic counted
resources on nodes (we are mostly using only ncpus, mem and vmem),
through virtualization, upto clustering (eg. job requesting infiniband
cannot be assigned nodes from different clusters).

Some of them are stable, some of them are work in progress.

I'm not pushing these patches into bugzilla because they are not yet
harmonized with the upstream semantics (like the previously mentioned
exclusive jobs).

We are planing to set up a testing version (on a small cluster) at the
end of the month with a full production environment (including P2P
scheduling) at the end of the year.

-- 
Mgr. Šimon Tóth


More information about the torquedev mailing list