[mapserver-users] Raster data performance

Wed Jul 28 07:09:53 PDT 2010

I don't know anything about your work flow, but if possible, you could add a pre-processing step where you use a script to read your text file and pre-classify the data.  You would end up with integers representing each class as 'cell' values.

Instead of :
EXPRESSION([pixel] >= -4 AND [pixel] < -3)

You could have: EXPRESSION "2"
Or
EXPRESSION /2/

Matching a single string should be the fastest expression to evaluate.

David.

From: mapserver-users-bounces at lists.osgeo.org [mailto:mapserver-users-bounces at lists.osgeo.org] On Behalf Of Tobias Weiß
Sent: Wednesday, July 28, 2010 8:25 AM
To: mapserver-users at lists.osgeo.org
Subject: Re: [mapserver-users] Raster data performance

Hi Alex,

I tried some of the performance tips you mentioned, e.g. converted the .txt to a tiled TIFF to optimize performance.  I can not see significant changes in rendering speed.

gdal_translate -co TILED=YES input.txt output.tif

The major brake still is the expression stuff. I really wonder what makes this task so time-consuming, it should not be more than a simple comparison.

Do you have experiences with this issue? Or do I have to live with this constraint?

Best, Tobi

===
Am 28.07.10 14:27, schrieb Alexandre Dube:
Hi Tobi,

  The following documentation topics about raster might help :

http://mapserver.org/input/raster.html#rasters-and-tile-indexing
http://mapserver.org/input/raster.html#raster-display-performance-tips

  and about expressions :

http://mapserver.org/mapfile/expressions.html

  The string comparison type is the fastest, so you could try define more classes but using string comparison to see if it gets faster.  Also, a good performance tip is to put your most frequent classes types on top first.  For example, if 80% of your data has "pixel = -3" then put it on top.

Kind regards,

Alexandre

On 10-07-28 04:54 AM, Tobias Weiß wrote:
Hello list,

this is my first email to this list. I am working with MapServer, OpenLayers & TileCache for almost 1 year now and the tools are really working great together, when everything is configured correctly ;-)

At the moment I´m facing a performance problem with raster data that serves as input data to render my map tiles. The spatial resolution of this raster data is 720 (rows) x 360 (cols) and it is stored in a simple .txt file (The size of the file is around 2 MB). I use TileCache to prerender maptiles for different zoom levels that are later displayed in a web application using OpenLayers.

I discovered, that the rendering speed of MapServer seems to mainly depend on how many different classes I define within my mapfile, meaning how many expressions MapServer has to evaluate during the rendering process. Without any classes MapServer the rendering speed is really fast, the more classes I use the slower it gets. With e.g. 10 different classes defining 10 different styles / pixel colors the rendering time for one tile can increase up to 60 sec.

CLASS

  EXPRESSION ([pixel] >= -4 AND [pixel] < -3)

    STYLE

        COLOR 69 117 180

   END

END

While working with shape files I haven´t had this problems and prerendering map tiles was really fast. With raster data the rendering time for one map and 5 different zoom levels can take up to several hours and almost overburdens my CPU.

Can anybody give me a hint how I could improve this?

Best, Tobi

_______________________________________________

mapserver-users mailing list

mapserver-users at lists.osgeo.org<mailto:mapserver-users at lists.osgeo.org>

http://lists.osgeo.org/mailman/listinfo/mapserver-users

--

Alexandre Dubé

Mapgears

www.mapgears.com<http://www.mapgears.com>

_______________________________________________

mapserver-users mailing list

mapserver-users at lists.osgeo.org<mailto:mapserver-users at lists.osgeo.org>

http://lists.osgeo.org/mailman/listinfo/mapserver-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/mapserver-users/attachments/20100728/326002da/attachment.htm>