[postgis-devel] PostGIS capability in the context of High Performance computing, grid computing

[maplabs at light42.com]

> From: Bryce L Nordgren <bnordgren at gmail.com>
> I've thought about this before, but it seems to me that in order to get much
> advantage out of the cloud, the geometric algorithms would have to be
> parallelized. Of course, I'm usually thinking in the space of very few
> queries each of which are difficult, rather than many simple queries. 
>

  Bryce - I bet you are a super smart guy, but I think you missed the 
mark here!
Brian Case and I have some some parallel experiments at Telascience in 
San Diego. 
We broke some problems down into parallel pieces, and then executed each one;
 recombine the results at the end.. Using stock postgres on linux as 
it is right now.. 

Most of the problems had to do with analysis on grids.. the grids can 
be square cells
or hex cells.. You could even break it down by region, if the regions 
dont overlap. 

and something else from last nights work.... :-)
Geocoding addresses.. a database is loaded into RAM, and I had four 
cores running
from the same shared memory allocation, each independently doing geocoding. 
It works perfectly!

  In Summary - some problems are decomposable into parallel chunks
With a little coordination, that can be done right now.. Rewriting the core
algorithms? I wouldn't hold my breath on that :-)

  best from Berkeley

==
Brian Hamlin
GeoCal
OSGeo California Chapter
415-717-4462 cell

  >
> On Wed, Aug 3, 2011 at 3:27 AM, jitendra <jituviju at gmail.com> wrote:
>
> >
> > Is there a document I can refer to in this regard
> > It is proposed to create a OpnGIS cloud on a powerful infrastructure ( see
> > www.garudaindia.org ). 
> > Wish to know if there are similar experiences any where?
> >
> > --
> > jitendra
> >
> > ________