[Mapserver-users] classification performance

Steve Lime steve.lime at dnr.state.mn.us
Thu Jul 22 12:17:20 EDT 2004


I don't know if it's expected but it's not all that suprising. I'd be
curious for a bit
more detail. For example:

 - how many features are in the bbox
 - how many features in the shapefile
 - average feature complexity (nverts)
 - distribution of classes in the bbox
 - the actual listing of expressions (layer definition) 

The order that classes are defined may have a siginificant impact on
performance,
especially in a case like yours. For example, let's say that the bulk
of the features
match class 27. That means that the first 26 expression tests all
failed. So it's not
just a case of one feature means one expression test, it can be much
worse.

Pre-classifying your date, either splitting into seperate shapefiles,
or applying the
expressions AHEAD of time (so you can use string expressions) can
really help.

Steve

>>> <bartvde at xs4all.nl> 7/22/2004 6:51:32 AM >>>
Hi list,

I have been doing some performance tests this morning on a shapefile
and a
tiled version of the same shapefile. The most astonishing outcome for
me
was however the amount of time a classifcation takes. I have a pretty
complex classification for my dataset with 28 classes, mostly using
regular expressions.

Here are the results of the test (Intel Xeon 2,4 Ghz, 1,5 Gb memory,
Windows 2000 Server), tests performed with Microsoft Web Application
Stress Tool on the same server, during 1 minute of continuous
requests:

1 Shapefile (1,6 Gb) "without" classification (1 class): 1.4 sec
Tiled shapefiles in 10x10 (=100) tiles without classification: 0.7 sec
1 Shapefile (1,6 Gb) with classification: 2.4 sec
Tiled shapefiles in 10x10 (=100) tiles with classification: 1.9 sec

So executing the classification on the dataset costs approximately
1-1.2
seconds, which is a more than reading the data in the case of a tiled
dataset. *** Is this to be expected? ***

bbox used: 160000,445000,170000,455000 (10.000 x 10.000 m)
image dimensions: 750x750 pixels
WMS protocol

Best regards,
Bart
_______________________________________________
Mapserver-users mailing list
Mapserver-users at lists.gis.umn.edu 
http://lists.gis.umn.edu/mailman/listinfo/mapserver-users



More information about the mapserver-users mailing list