Raster tile indexing in PostGIS

[Thy, Kristian]

Frank et al,
 
First, your comments and thoughts are much appreciated. I am trying to form a picture for the infrastructure we are going to build to serve a high-performance WMS with fairly static data, so I guess shapefiles sound more efficient than the database access. But one thing came to mind while buying my groceries - how do shapefiles perform under increasing load on the server? I.e. is it possible for several instances of MapServer to read concurrently from the same shapefile and spatial index, or will the database help on performance in this scenario? I know it's dependent on the nature of my data and the number of concurrent users, but if it's possible to give some sweeping comments on this issue I'm very interested in hearing your opinions.
 
best,
Kristian

________________________________

Fra: UMN MapServer Users List på vegne af Frank Warmerdam
Sendt: on 22-11-2006 16:56
Til: MAPSERVER-USERS at LISTS.UMN.EDU
Emne: Re: [UMN_MAPSERVER-USERS] Raster tile indexing in PostGIS



Thy, Kristian wrote:
> From: percy [percyd at pdx.edu]
>> My understanding is that shapefiles with a spatial index
>> (qix, generated with the shptree utility in mapserver) are
>> actually faster than postgis access to the same data.
>
> From: Fawcett, David [David.Fawcett at state.mn.us]
>> If you are looking for performance, I am guessing that you can
>> do better with spatially-indexed shapefiles.  Have you created
>> quadtree indexes for your tileindex shapefile?  Take a look at
>> shptree:
> http://mapserver.gis.umn.edu/docs/reference/utilityreference/shptree
>
> Oh my.
>
> Is this the general consensus? I.e. if I'm just using MapServer
> for displaying data and not using any of the spatial query
> stuff from PostGIS, I'm better off (performance-wise) with plain
> old shapefiles?

Kristian,

For simple cases the consensus is that shapefiles with a spatial
index are faster in mapserver than postgis is.  There are exceptions
of course, such as when you want to do attributes based queries which
cannot be indexed in shapefiles with mapserver.

> And if so, is the fact that ArcSDE is seemingly faster than
> plain old shapefiles just another ESRI license selling
> point to get people to shell out money for SDE? ;-)

It is hard to imagine that SDE is really faster for simple operations
than a shapefile unless they have made no effort to make shapefiles
fast.  I would presume the main selling points of SDE are about
shared access in a workgroup, "managability", and the ability to
do complicated RDBMS'ish things (like joins) efficiently.

> If yes, it's going to be hard to convince my bosses that we don't
> need the database, but I'll try :-)

You don't have to try too hard.  Postgres+postgis is still quite fast.
But if you are very performance sensitive and the other attributes aren't
important to you then shapefiles will generally be better for mapserver
feature data (and tileindexes).

Don't forget the spatial index though!

Best regards,
--
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org <http://osgeo.org/> 


This message has been scanned for viruses by MailControl - (see http://bluepages.wsatkins.co.uk/?4318150)




This email and any attached files are confidential and copyright protected. If you are not the addressee, any dissemination of this communication is strictly prohibited. Unless otherwise expressly agreed in writing, nothing stated in this communication shall be legally binding.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/mapserver-users/attachments/20061122/1cc961ea/attachment.html