[Beowulf] Re: SGE + policy

[Robert G. Brown]

On Thu, 27 May 2004, Andrew Piskorski wrote:

> On Thu, May 27, 2004 at 12:01:09PM -0700, beowulf-request at beowulf.org wrote:
> 
> > Date: Thu, 27 May 2004 10:19:00 -0400 (EDT)
> > From: "Robert G. Brown" <rgb at phy.duke.edu>
> 
> > Primary tasks:  matlab and stata jobs, run either interactively/remote
> > or (more likely) in batch mode.  Jobs include both "short" jobs that
> 
> > Constraint:  matlab requires a license managed by a license manager.
> 
> Have they considered running Octave rather than Matlab?  Depending on
> what special-purpose Matlab add-on modules they're using, that might
> or might not be practical, but if it is, using Octave should reduce
> their cluster and license hassles.  (I've never used either of the two
> dialects myself though, so have them ask someone who really knows...)

They aren't quite there yet.  First we build the pilot cluster and make
it "work" for their existing work cycle using the tools that they are
used to, THEN we start them experimenting with cheaper/better
alternatives.

The cost scaling of matlab is actually a major strike against it
anywhere, IMO.  Even if one grants that it is a fabulous program and
much better than maple, mathematica, octave (all of which live in
similar spaces with widely varying costs) it is hard to manage in a
cluster environment.

This probably doesn't worry them, because in one sense anybody who uses
matlab for doing "serious" computations probably needs therapy, and
clusters are often build for doing serious computations.  However, this
particular cluster is being built as much to support many students doing
less serious computations as it is to support work that really should
probably be done in c.* or f.*, and in a lot of cases the real expense
is teaching people with little or no real programming experience how to
convert from a garbage-tolerant interactive environment to a compiled
language.

   rgb

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu