[mapserver-users] HOWTO Split Large SHP Files

[Paul Ramsey]

My number one performance hint for mapserver: never render more data
than you have to. If your data is more detailed than is showing up in
the rendered map, you're wasting cycles. Create a thinned data set, with
less features, or with vertex-reduced features, or both. Or use an
existing "higher level" data set. Use the 'GROUP' flag to group all your
logically identical layers (roads, streams, lakes, whatever) into a
single logical layer which you can then toggle in your interface. Scale
dependant layers are your friend, always.

Luca Pescatore wrote:
> 
> >Correct me if I am wrong here Luca, but your performance problems occur
> >at MAXIMUM ZOOM? You have described massive shape files of 300Mb+ worth
> >of roads. At maximum zoom, you should not be rendering all the roads in
> >your dataset.
> >It's a waste of time to render every road in your dataset,
> >they will be mostly invisible. You need extract the important roads
> >(highways, etc) and only display those roads at MAXIMUM zoom. Use the
> >max and min scale parameters to flip from the small dataset to the
> >larger dataset at an appropriate time. Build shptree's on your big
> >datasets so that at smaller zooms you get good performance.
> Maximum zoom loads all roads, at other levels i load only major
> highways or secondary highways.
> 
> shptree have improved performances, i will put in this list some test for my
> machine, i think that will be useful in future for who will implement a
> mapserver.
> 
> Next Week i will put another 512 Mb of ram and i will re-test everything.
> 
> So i will get some test info without shptree indexing (512 Mb RAM), with
> shptree indexing (512 Mb RAM),
> without shptree indexing (1 Gb), with shptree indexing (1 Gb), without
> shptree indexing
> (1 Gb) and small files and without shptree indexing (1 Gb) and small files.
> 
> (I think that may be a very nice case history... kewl! :) )
> I will put on my site some info about that.
> 
> Best Regards,
>          Luca