Hi,

We are working with a bit similar data right now. Myself I do not think that 
drawing all the points is the best way for showing their distribution if points 
are so dense that the rendered dots start to overlap. Let's say the  symbol is 
6 pixel wide red dot. If there are observations in the data 6 pixel apart each 
other the map will be painted all red. User will see no difference with places 
where observations are one 1 pixel apart. We aim at making heathmaps from our 
observation data instead. Grouping the amount of observations and draw the 
groups with variable size symbols was another approach but heatmaps suit better 
with our data because we want also visualize the amount of damage from the 
attribute data.

I did do some experiments a few years ago and my conclusion was that there are 
no need to draw more than a few thousand features (parcels in that case) on a 
screen because there are not really enogh many pixels to show them all and 
painting same pixels many times is just waste of time. However, the sample must 
show a similar distribution than the original data. If distribution is uneven 
and there can be, let's say, a big accident in a very isolated place and it 
should show on a map at any scale then it would be bad if it drops out from the 
sample.

I can make a test next week with our Oracle next week. Hypothesis is that if 
table has 50000 rows then rendering everything is fast, but if I render a table 
with 150000 rows then it will take more than three times longer.

-Jukka Rahkonen-



cheesybiscuits wrote:

> Yukka: unfortunately yes, at this point it is necessary to draw all the
points. I have to show their distribution so I can't hide some and show
others as it would misrepresent the overall shape. As their locations
frequently change it is not feasible to calculate any of this in advance and
calculating on the fly would presumably incur a higher performance cost.

> However, even if I could reduce the number of points displayed the
performance question - and whether this is typical - still stands.

> cholmes: I might be able to try this, or test with a Shapefile, but Oracle
is already in place as the operational database and there is no scope to
change it. Do you have any suggestions on where I might find more
information about these bottlenecks?

> Also I'm a bit confused about how Oracle would affect rendering performance
if what I think it happening is true - after Oracle finds all the data it
takes GeoServer 5 seconds to display it and Oracle does very little until
the connection is closed - am I missing something here?

> Thanks both for your input



--
View this message in context: 
http://osgeo-org.1560.n6.nabble.com/5-seconds-to-render-130-000-points-typical-GeoServer-WMS-performance-tp4995984p4996023.html
Sent from the GeoServer - User mailing list archive at Nabble.com.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Geoserver-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geoserver-users

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Geoserver-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geoserver-users

Reply via email to