Multiprocessor/cluster versions of GRASS

[Rich Shepard]

  After a couple of days of investigation and e-mail discussions with the
experts, I can now report that GRASS, SWAT and the other neat tools
available to us can be re-compiled to take advantage of parallel processing.

  There may be other vendors out there, but I know several of the folks at
The Portland Group (http://www.pgroup.com), and they've been very helpful.
Their high-performance, aggressive-parallelization compilers are widely used
in the supercomputing sector.

  The brief answer is that GRASS, SWAT and the other models can run on
multiple processors or clusters. TPG's toolkits will analyze the code and
determine what needs to run on a single processor and what can run on
multiple ones. They have both FORTRAN (77 and 90) and C/C++ compilers
available.

  Although this is just something I'm considering for much later this year,
I'm looking at an EZ Parallel Supercomputer 2796
<http://ezcp.cjb.net/supercomputers.html> with its 6 Celeron 466MHz
processors. The price is reasonable (currently US$6,500); the next step up
is a better value: 16 processors and a larger hard drive, but the price is
US$15,000. Anyway, these are beowulf clusters; a linux supercompter.

  If we develop the business where we are running GRASS/SWAT models and
simulations on a regular basis, I'd like to work the the Portland Group
folks to produce a multiprocessor version.

  So, it can be done. I wonder how long it would take, Bruce, to run HUMUS
on the US$20,000 Supercomputer 8800 with its 16 PIII/550MHz cpus and 2G RAM?
Wouldn't that be an interesting comparison?

Rich

Dr. Richard B. Shepard, President

                       Applied Ecosystem Services, Inc. (TM)
              Making environmentally-responsible mining happen. (SM)         
                       --------------------------------
            2404 SW 22nd Street | Troutdale, OR 97060-1247 | U.S.A.
 + 1 503-667-4517 (voice) | + 1 503-667-8863 (fax) | rshepard at appl-ecosys.com