[mapserver-users] expected speeds.

[Sam Ingarfield]

Hi;

I've just been through a bit of an optimisation process on my server and was
wondering whether the speeds that I am obtaining are somewhat inline with
what could be expected with the hardware/datasets that are in use.

Firstly: Hardware.
2.2Ghz Xeon (Single Core, but with hyperthreading. not that hyperthreading
helps much.)
1.2Gb Memory coupled with over 2Gb of swap (was filling up memory and
original swap realllly fast so swap was expanded)
Data being served off an array of three ultra320 15k scsi drives in a raid-5
config.

Software.
OS: Debian
Mapserver: Nighly build as of 6/5/09, in fastcgi mode
Data source: Postgis database

Data:
Coastline Layer (9 Polygons with a stupid number of vertices,  this is a
known slow point due to toast table problems. or at least we think it is)
Islands Layer (~9000 polygons, only ~4mb)
Internal Waterbodies layer (~11000 polygons, plus ~1.2m lines, 800mb all up)

Performance:
On a moderate zoom level, about 2s per 256x256 tile.
On more zoomed out levels, process times out before completing. and yes,
some generalisation needs to take place at further out zoomage but thats not
sorted yet.
If i drop the coastline layer out (the toast problem one), speed ups to like
80-100ms/tile


So, watchathink?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/mapserver-users/attachments/20090506/b5d9446b/attachment.html