[mapserver-users] Displaying Huge Data on Mapserver

[Stephen Woodbridge]

Frank Warmerdam wrote:
> Rishikant Lal wrote:
>> Dear Friends,
>> I have a problem. I am trying to display nearly 10 Million Rows of 
>> Point feature on Map Server. I have imported the Data on Post GIS, 
>> Created GistIndex and even used Fast CGI. Features Display very 
>> quickly without Labels. Problem occurs if I try to label them in Layer 
>> definition.
>> Is there a way I can Label Features in MapServer, without affecting 
>> the Map Display Speed.
> 
> Rishikant,
> 
> The literal answer is no, you can't do more work without some impact on
> performance.
> 
> How many points are you trying to label?
> 
> My understanding is that label collision detection in MapServer can get
> very expensive when a lot of of labels are candidates for display.  If
> you attempt to display an area in which 10000 points appear, it will be
> very expensive even if after label collision detection they are weeded
> down to a much smaller number.
> 
> If this is the issue, then you need to consider ways of restricting the
> labelling to more reasonable numbers of points.
> 
> Best regards,

Rishikant,

As Frank says label collision detection is very expensive and is close 
to an O(n*n) problem with respect to the number of candidates. So 
ideally you want to be able to tag some points that are more important 
than others or create another layer where you have clustered points in 
to some smaller number of of cluster points and display that when you 
are zoomed out and use that until you zoom in enough to spatially filter 
the points to the map viewport.

If all points have equal display value, then you might add a column and 
assign a random number between 1 and 10 to each row, then you can filter 
based on that. To get 10% of the points, "where number=1" or 30% of the 
points "where number<=3", etc. In index may or may not help.

-Steve W