WMS performance shootout results

[Michael Tafel]

Hello list.

I thought people would be interested in the results from the WMS performance shootout that occurred at FOSS4g2009 in Sydney this past week.

Originally was supposed to be AGS v Geoserver v Mapserver but ESRI dropped out at the last minute.

Results:

http://tinyurl.com/yg4ko3w

Cheers,

Mike Tafel

[Carl Reed - OGC]

Mike -

Great content! Thanks to all who made the effort in this activity. I
will be sharing with the OGC.

Regards

Carl

[Randy George]

Too bad ESRI didn't stay in the shootout.

"PostGIS performs roughly 5% better than Oracle on higher loads"
"SDE seems to introduce no overhead compared to direct Oracle access for
GeoServer, and slight overhead for MapServer"

Looks like putting PostGIS behind SDE could offer an improvement and putting
GeoServer/MapServer in front of SDE would very likely help OGC service
performance. Can't really tell unless ESRI participates. SDE shrinks to
analysis/data creation?

Next year it would be interesting if SQL Server Spatial could be added to
the mix.

Randy

[Tyler Erickson]

I had heard that ESRI dropped out due to 'lack of resources' which may
have been a valid reason since they announced that they would join in
the shootout rather late in the game. On the other hand, Andrea and
Jeff said they ran their benchmark tests on weekends during the month
prior to FOSS4G, so maybe it is more of a comparison of how willing
open source vs. proprietary software developers are willing to donate
their free time.

Because next year's WMS shootout in Barcelona has already been
announced, I assume that ESRI will be able to allocate enough
'resources' to join in, if they are truly interested in doing so.

The Oracle vs. PostGIS results are interesting, but the difference may
have been partly due to the developers experience in configuring the
respective databases. During the presentation they didn't really
discuss how much effort was made in configuring the databases, or
whether they just used an out-of-the-box installation of the
databases.

Also I had heard that publishing benchmark results is in violation of
Oracle's end-user license, so does this mean that this was a one-time
test using the Oracle backend (or has Oracle changed their terms)?

- Tyler

[Randy George]

I had heard that about Oracle as well. Not sure if it is just legend, but it
doesn't seem contrary to the Oracle persona.

Maybe Oracle could send some of their people to Barcelona to tweak the
Oracle benchmark.
Interesting to see Amazon RDS built on MySQL which is an Oracle/SUN open
source? I still haven't figured how open source can be owned?

So benchmarks to see:
PostgreSQL/PostGIS open source
MySQL sort of open source (especially interesting to see it on an
Amazon RDS with all tables in memory :)
Microsoft SQL Server
Oracle
IBM DB2

If anyone missed Paul Ramseys talk at FOSS4G "Beware of Geeks Bearing Gifts"
you should take a few minutes. It is an insightful look at FOSS
http://www.youtube.com/watch?v=zB_a28vBtBk