TL;DR: when using only 2 or 4 out of 8 total cores, the virtualized
gizmo manages better memory locality and better cache usage because
of less motion of processes to different cores. In HPC, we generally
lock processes to cores, and they should done this, too.

Even shorter: Pilot error, nothing to see here.

On Wed, Aug 14, 2013 at 03:10:56PM -0400, Douglas Eadline wrote:
> 
> "When Virtualized HPC Speeds Past Bare Metal"
> 
>   http://insidehpc.com/2013/08/14/when-virtualized-hpc-speeds-past-bare-metal
> 
> Darn, I wish I had that secret virtual sauce, but I am stuck
> with that slow inefficient bare metal. However, I am going to
> look into virtualizing the virtualized layer
> so that I can "speed past" simple "Virtualized HPC."
> 
> 
> --
> Doug
> 
> -- 
> Mailscanner: Clean
> 
> _______________________________________________
> 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
_______________________________________________
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

Reply via email to