VS: [UMN_MAPSERVER-USERS] Improving Performance... {Scanned}

Rahkonen Jukka Jukka.Rahkonen at MMMTIKE.FI
Tue Mar 21 06:32:37 EST 2006


Jelmer Wrote

>Rahkonen Jukka wrote:
>> Hi Jelmer,
>
>> I made a quick test with one shapefile of size 580 MB containing 1.2
>> million polygons. Shptree (from FW Tools 0.9.8) with default
>> parameters took 2 min 30 secs to run with 3 GHz Pentium and Windows
>> XP and memory usage grew not more than 200-300 MB during the process.
>> Of my 1 GB I had all the time more than 500 free. The idle usage
>> seems to be around 200 MB.

>Hmm, for some reason the file that I mentioned earlyer DOES work. I
>probably used a different file for testing, though I can't remember
>which one.

>Can you comfirm if the SHP+index is faster than the .TAB file?


I fear it is not that easy to say, but a bit better study should be planned. It is easy to set up a layer with 1.2 million polygons and ask Mapserver to render it to a window of 800 by 800 pixels. I have also it and Mapserver failed. I know as well that it can draw me something like 200000 polygons without a trouble, but I think neither of these tests do not have any real meaning despite that it is nice to to know where the limits are.  Even in the latter case there are only about 3 screen pixels on average for drawing each polygon and it means that the user for sure cannot see them all. The screen would look about the same with 20000 polygons only, and with 2000 polygons the user might see some usable information as well.

I believe that some usable numbers about the differences in performance could be achieved if you could first try to find the largest realistic view on the screen and then perform for example 50 ramdom map searches without changing the scale. Of course you should have no other layers in the map but only either .TAB or .SHP layer.  I believe both alternatives are rather fast if you draw only for example 5000-10000 features at the time, but this is just my guess.

-Jukka Rahkonen-

 



More information about the mapserver-users mailing list