[Bioclusters] cluster or SMP

[Joe Landman]

Robert Olson wrote:

[...]

> It takes a fair bit of work and understanding of local vs remote file
> system issues to make the cluster work efficiently. However, we have
> seen *immediate* increases in productivity for the folks that started
> using the SMP boxes; the increase in amount of memory and the fact that
> it is trivial to efficiently use the multiple processors made a world of
> difference.

Anecdotally, in a different market, we build 4 and 8 core desktop
machines for customers with 8->32 GB ram, as frontends for their
clusters.  We find that users like running smaller jobs on these
machines, and submit larger ones to the clusters.  These customers are
even running their Windows desktops virtually using VMWare sessions.

I am working on an 8-core desktop unit now, doing some additional
testing with it.  Small SMP units like this are really nice to have to
do work on IMO.

[...]

> As far as network goes; we've got 2TB of local disk on each of these
> machines (3 750G drives), so we have essentially zero network use for
> the work being done on them at the moment...

This is similar to our design.  4x500 GB units in a RAID10.  Redundancy
and speed.



-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
       http://jackrabbit.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 866 888 3112
cell : +1 734 612 4615