OpenGeo

From: aaime

Subject: WPS and spatial databases

Date: Mar 15, 2010 04:37 AM

Hi,
during the weekend I was thinking about making spatial
processes that are backed by a spatial database.

Say, you want to overlay two layers, you put them into
PostGIS, run a query that does the overlay, grab
the results, cleanup.
PostGIS is clearly quite fast at the overlay job and
does scale to high volumes of data.

But there are two catches:
- the code has to move all the data from and to PostGIS,
   which is working in another process
- PostGIS has to be there in the first place

It also has implications on the hardware of choice.
Usually database servers are not setup to do intensive
processing, they are more geared towards peak I/O
(so normally they are not the best choice to run
  complex algorithms fast).

If we used a spatialized H2 instead it would be embedded
and we can make sure it's there as we use it embedded.
However I'm wondering about what kind of spatial operations
we can do there.
Does H2 provide all JTS methods against the stored geometries?

What about aggregate functions?
Trying to understand how much functionality we give up
if we created spatial processes on top of H2 instead of
PostGIS

Cheers
Andrea

-- 
Andrea Aime
OpenGeo - http://opengeo.org
Expert service straight from the developers.

Thread Outline:

• Re: WPS and spatial databases — jdeolive

From: jdeolive

Subject: Re: WPS and spatial databases

Date: Mar 15, 2010 11:29 AM

On 3/15/10 4:37 AM, Andrea Aime wrote:
> Hi,
> during the weekend I was thinking about making spatial
> processes that are backed by a spatial database.
>
> Say, you want to overlay two layers, you put them into
> PostGIS, run a query that does the overlay, grab
> the results, cleanup.
> PostGIS is clearly quite fast at the overlay job and
> does scale to high volumes of data.
>
> But there are two catches:
> - the code has to move all the data from and to PostGIS,
> which is working in another process
> - PostGIS has to be there in the first place
>
> It also has implications on the hardware of choice.
> Usually database servers are not setup to do intensive
> processing, they are more geared towards peak I/O
> (so normally they are not the best choice to run
> complex algorithms fast).
>
> If we used a spatialized H2 instead it would be embedded
> and we can make sure it's there as we use it embedded.
> However I'm wondering about what kind of spatial operations
> we can do there.
> Does H2 provide all JTS methods against the stored geometries?
>
> What about aggregate functions?
> Trying to understand how much functionality we give up
> if we created spatial processes on top of H2 instead of
> PostGIS
I have tried to implement most of what is possible with JTS and most of 
what is available in postgis. You can find a list of the all the 
supported functions here:

http://github.com/jdeolive/geodb/blob/master/core/src/main/java/geodb/GeoDB.java

There is not much there currently in terms of aggregate functions 
although they would be pretty straight forward to implement.

One of the next steps also is to use prepared geometries for operations 
over a result set.

-Justin

>
> Cheers
> Andrea
>


-- 
Justin Deoliveira
OpenGeo - http://opengeo.org
Enterprise support for open source geospatial.