[ka-Map-users] KaMap performance tuning

[Paul Spencer]

Hi Jeff,

this is an interesting analysis.  For the record, I always thought  
properly tuned shapefiles (tiled, indexed, sorted) were faster than  
any other format.

I suspect that the actual map rendering time relative to the other  
processes involved (as you said, Apache, PHP and also the PHP code  
that slices the resulting image, plus using wget invokes a network  
stack and various other communications pieces) is fairly  
insignificant in your case (and perhaps in general).  I don't really  
know of a good way to isolate the various factors.  Well, you could  
look into using xdebug with PHP and invoking tile.php directly to see  
what the overhead is.

If this is the case, it would be awesome to have a custom libmap  
executable that could take the place of tile.php - essentially a C  
program (preferably that could run either on command line or as cgi)  
that would create tiles built on top of the mapserver C API (such as  
it is).  Given the simplicity of the task, any scripted approach  
(such as tile.php) is going to be less than optimal.

Any volunteers?

Paul


On 21-May-07, at 3:25 PM, Jeff Dege wrote:

> We've been using KaMap with Mapserver and OpenLayers, with some  
> success.
> We've recently begin looking at using PostGIS instead of shapefiles,
> hoping to getter performance when generating maps that haven't yet  
> been
> tiled.
>
> So we loaded our shapefiles into PostGIS tables, and created a new map
> file that was a copy of the first, except that it got its map data  
> from
> the PostGIS tables.
>
> Our first comparison was using shp2img, which showed a dramatic  
> decrease
> in the time needed to generate maps - with our data, PostGIS was eight
> to ten times faster.
>
> So we cleared out our tile cache, pointed KaMap at the PostGIS map  
> file,
> and saw very little difference.
>
> So we created a batch file that would download a thousand different
> tiles, using wget to make direct calls against tile.php, and timed how
> long it took to run, with the shapefile and the PostGIS map files.   
> The
> PostGIS test was only five percent that the shapefile test.
>
> So we created yet another batch file that used wget to make direct  
> calls
> against mapserver.exe, running in CGI mode.  When generating maps from
> PostGIS data, mapserver was eight times faster than against  
> shapefiles.
>
> So now the question:  If the image creation is so much faster, when
> accessed through shp2img, and when accessed through mapserver.exe, why
> is that speed improvement not visible when loading tiles through  
> kamap's
> tile.php?
>
> It's clear to me that there must be some bottleneck, reducing
> performance.  My guess would be in Apache or PHP configuration.
>
> Does anyone know how I can measure what time is being spent doing  
> what,
> in this area?  What configuration options I should look at, to try to
> improve performance?
>
> Thanks.
>
>
> _______________________________________________
> ka-Map-users mailing list
> ka-Map-users at lists.maptools.org
> http://lists.maptools.org/mailman/listinfo/ka-map-users

+-----------------------------------------------------------------+
|Paul Spencer                          pspencer at dmsolutions.ca    |
+-----------------------------------------------------------------+
|Chief Technology Officer                                         |
|DM Solutions Group Inc                http://www.dmsolutions.ca/ |
+-----------------------------------------------------------------+