Hi,
I guess that classification has a big effect in your case. I have been reading
that string comparisons are faster than logical ones. How about to have a try
by classifying your shapefile and writing the classes directly to a new
attribute field?
This is my scale dependent group, simplified and spiced with some typos, I
fear.
LAYER
GROUP "Parcels"
NAME "all_parcels"
TYPE POLYGON
STATUS ON
DATA "parcels_2008"
UNITS METERS
MAXSCALEDENOM 100000
CLASS
NAME "parcel_class"
STYLE
OUTLINECOLOR 250 0 180
WIDTH 2
END
END
METADATA
"ows_include_items" "AREA, PERIMETER"
"wms_title" "all_parcels"
"wms_srs" "EPSG:2393 EPSG:3067 EPSG:4326 EPSG:900913"
END
PROJECTION
"init=epsg:2393"
END
END
LAYER
GROUP "Parcels"
NAME "ten_percent"
TYPE POLYGON
STATUS ON
DATA "parcels_2008_0"
UNITS METERS
MAXSCALEDENOM 1000000
MINSCALEDENOM 100000
CLASS
NAME "parcel_class"
STYLE
OUTLINECOLOR 250 0 180
WIDTH 2
END
END
METADATA
"ows_include_items" "AREA, PERIMETER"
"wms_title" "ten_percent"
"wms_srs" "EPSG:2393 EPSG:3067 EPSG:4326 EPSG:900913"
END
PROJECTION
"init=epsg:2393"
END
END
LAYER
GROUP "Parcels"
NAME "one_percent"
TYPE POLYGON
STATUS ON
DATA "parcels_2008_00"
UNITS METERS
MINSCALEDENOM 1000000
CLASS
NAME "parcel_class"
STYLE
# COLOR 240 240 0
OUTLINECOLOR 250 0 180
WIDTH 2
END
END
METADATA
"ows_include_items" "AREA, PERIMETER"
"wms_title" "one_percent"
"wms_srs" "EPSG:2393 EPSG:3067 EPSG:4326 EPSG:900913"
END
PROJECTION
"init=epsg:2393"
END
END
LAYER
-Jukka-
________________________________
Lähettäjä: Chris Jackson [mailto:webturt...@gmail.com]
Lähetetty: 3. syyskuuta 2010 12:13
Vastaanottaja: Rahkonen Jukka
Kopio: mapserver-users@lists.osgeo.org
Aihe: Re: [mapserver-users] Display performance
Hi Jukka
Thanks for the response and the great investigative work!
My log entry for drawing the layer is - [Fri Sep 03 10:01:25
2010].658000 msDrawMap(): Layer 1 (meanspringpeakflow), 24.500s. :o(
I do wonder if my Mapfile layer definition is just v.poor (admittedly
no expert!) so have taken an excerpt of the layer defintion (there are about 15
classes in all) - views anyone, should I be loading it a different way?
Also Jukka, could I get a sample of your scale dependent WMS GROUP
layer map code - as it sounds like a great idea maybe for this but also some
v.hi-res model grids I am thinking about.
Thanks again, much appreciated!
Chris
LAYER
NAME "meanspringpeakflow"
TYPE polygon
DEBUG 5
DATA "Tide"
TEMPLATE void
PROJECTION
"+proj=utm +zone=31 +ellps=GRS80 +units=m +no_defs"
END
METADATA
"DESCRIPTION" "Mean Spring Peak Flow"
"RESULT_FIELDS" "ID,DISTANCE,DEPTH,MEAN_SP_PC"
"RESULT_HEADERS" "ID,Minimum Distance (m),Average Depth (m),Mean
Spring Peak Flow (m/s)"
"ows_title" "meanspringpeakflow"
"RESULT_HYPERLINK" "ID|| Load graphing tool"
END # Metadata
CLASS
NAME '> 4.00 (m/s) '
EXPRESSION ([MEAN_SP_PC] >= 4 AND [MEAN_SP_PC] < 4.5)
STYLE
COLOR 135 99 64
END #end style
END # end class
CLASS
NAME '3.51 - 4.00'
EXPRESSION ([MEAN_SP_PC] >= 3.5 AND [MEAN_SP_PC] < 4)
STYLE
COLOR 158 126 63
END #end style
END # end class
etc etc
On 3 September 2010 09:04, Rahkonen Jukka <jukka.rahko...@mmmtike.fi>
wrote:
Hi,
I took some numbers to compare with.
Material and methods
Polygon layer, rather simple polygons with few vertises,
Feature Count: 117383
DEBUG 5 in mapfile
WMS client asks for the whole layer
Mapserver 5.2.1 (CGI) on not so fast Windows computer
Results
First request: rendering time taken from the log file are
between 3.5 and 4.5 seconds
Following request, BBOX is changing but all the polygons are
drawn though: rendering times 1.5 - 2.5 seconds
Conclusions and discussion
At first one might think that my server is faster. But this
test is not controlled at all because we are not using the same shapefiles and
we can't say so. What we can say that a couple of hundred of thousand polygons
can be drawn faster.
However, it does not really make sense to render an image this
way if it should be fast. If the screen has 1000 by 1000 pixels it makes a
million pixels together. For 200000 polygons it makes 5 pixels per polygon.
You can simplify your polygon geometries pretty much before anybody can see the
differense. If the polygons are spread evenly nobody can even see the
difference if there are 20000 or 200000 polygons on the screen.
We have one polygon layer with about million polygons and for
that I made two simplified layers which contain only 1 percent and 10 percent
of all the polygons. Those and the original shapefile are put into the same
scale dependent WMS GROUP layer. This group is pretty fast at any scale. The 1
percent layer is perhaps a bit too sparse but I have been too lazy to have a
try with 2 or 5 percent samples.
-Jukka Rahkonen-
_______________________________________________
mapserver-users mailing list
mapserver-users@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/mapserver-users
