r.surf.contour

[Will Mitchell]

I have noticed that the biggest performance factor is the choice 
of the location for the XDRIVER pipe.  If the pipe is across the
net (say by an NFS mount) you will have a major performance 
degradation.

Try to have the XDRIVER use a local filesystem.




   Sender: lists-owner at max.cecer.army.mil@comsearch.com Reply-To:
   grassu-list at max.cecer.army.mil@comsearch.com Precedence: Bulk Date:
   Sat, 12 Sep 92 10:52:33 EDT From: Neel Smith
   <nsmith at yogi.bowdoin.edu@comsearch.com> X-Mailer: ELM [version 2.3
   PL0]

   I'm trying to run r.surf.contour over a large rasterized topo map
   to create a DEM.

   I've had to slice the area up and run r.surf.contour over little
   pieces of it, then use r.patch to put them back together.

   I've simultaneously started jobs running from various grass users,
   and have noticed that if I start the job from the GRASS server, it
   runs far faster than if I start the job from a workstation.  Ex:
   serving froma VAX 8600, a 400 * 400 pixel job takes a few hours; on
   a DECstation 5000, the same job takes serveral days!

   This seems far greater than a difference in cpus could explain.
   DOes the client/server architecture of grass really impose that
   much of a burden on networked workstations?  Or have I missed
   something?

   Thanks for any advice.  I've got a whole lot more of these jobs to
   run, and if anyone can suggest how to keep the running time, I'd
   appreciate any advice!

   Neel Smith nsmith at polar.bowdoin.edu