Re: Postgres scalability and performance on windows

["J. Andrew Rogers" <jrogers(at)neopolitan(dot)com>]

On Nov 28, 2006, at 8:24 AM, Tom Lane wrote:
> "Gopal" <gopal(at)getmapping(dot)com> writes:
> > This is the query and the schema....
> > ...
> >                         select
> > sum(area(intersection(snaptogrid(chunkgeometry,0.00000001),
> >                         GeometryFromText('POLYGON((-0.140030845589332
> > 50.8208343077265,-0.138958398039148  
> > 50.8478005422809,-0.0963639712296823
> > 50.8471133071392,-0.0974609286275892  
> > 50.8201477285483,-0.140030845589332
> > 50.8208343077265))',4326))) * 100/ (0.00114901195862628)) as
> > percentCover,
>
> So evidently area(intersection(snaptogrid(...))) takes about 300
> microsec per row.  The PostGIS hackers would have to comment on  
> whether
> that seems out-of-line or not, and whether you can make it faster.


This is consistent with the typical cost for GIS geometry ops -- they  
are relatively expensive.  When running queries against PostGIS  
fields for our apps, about half the CPU time will be spent inside the  
geometry ops.  Fortunately, there is significant opportunity for  
improvement in the performance of the underlying code if anyone found  
the time to optimize (and uglify) it for raw speed.


Cheers,

J. Andrew Rogers