We are starting of a series of benchmarks with the new version of IWS. A long with this, were are also doing some testing with other server software, to compare the performance.
I will be publishing the results on this blog over the next two weeks.
For the first test, a loop was setup off through a pre-built CSV file with 2000 sample BBOX' of varying scale. Each HTTP request is randomised between 256x256px up to 1024x768px to reproduce real world settings. A 150 users were simulated.
This is the result of my first, initial test.
* Server X and Server Y represents what some would say "competitive technologies" and I was asked by the higher-ups not to publish the results. I compromized by putting in aliases instead....
Comments:
Whilst IWS is multiple times faster than the other servers, I think that i can get more out of it. The configuration of this performance options is slightly different in the new version, and I am still geting the hang of it.
Test Conditions:
Client
Windows Vista SP2
Intel Core 2 Duo T9600 @ 2.8ghz
4gb RAM
100mbit network to server
JMeter 2.3.4
JDK 1.6.0_16
Server
Windows Server 2008 x64
IIS Web server v7
Intel Xeon E5410 @ 2.33ghz (2x4 core)
16gb RAM
Data HDD Read speed: 160MB/sec
Software
Geoserver 2.0 stable.
- Windows Service Jetty binary installer or Tomcat 6.0.20
- APR 1.1.6, JVM options applied as per wiki http://geoserver.org/display/GEOSDOC/2.6+GeoServer+in+Production+Environment
- Reconfigured server.xml with 8 threads
- Production logging enabled
- Resource limit set to default 60secs
- GDAL 1.4 extension
- windows32-imageio-ext-installer-gdal-mrsid-ecw-1.0.4
Mapserver 5.4.2
- OSGEO4W installer
- No Mapserver 5.6 binary available as at 5/11/09
- GDAL 1.6
- Apache configured for fcgi on port 85
- Configured with 8 worker processes
ECW
- world-topo-bathy-200406-3x86400x43200.ecw (3 bands, 86400pixels x43200 pixels )
- 263mb compressed. About 10 GB uncompressed.
- WGS84