[mapserver-users] Mapserver speed/load test

[Seth G]

Hi all,

I gave the presentation in Bonn - I've uploaded the slides to 
http://www.slideshare.net/CompassInformatics/stress-testing-and-analysing-mapserver-performance 


I was testing performance for my own use case, to see if moving to Linux 
would give our system a performance boost. All our data comes from MS 
SQL Server, and using the FreeTDS driver on Linux compared to the native 
MSSQL driver on Windows appears to cause our setup to run slower overall 
on Linux.

There should be no conclusion that Windows is faster than Linux for 
other data sources, however it does appear that there is no huge 
difference between the 2 operating systems that there appeared to be 
several years ago (maybe due to improvements in the MSVC runtimes).

I'd recommend people to test their own setups and data, as any generic 
conclusions are almost impossible to make. I found the Locust load 
testing tool to be easy to work with as a Python developer. A sample 
task set can be seen at 
https://gist.github.com/geographika/55b8f4c6ae721ce8f7986c3da790bcf6

Regards,

Seth

On 05/10/2016 20:29, Märt Reose wrote:
> Age old battle of windows vs linux...
>
> That presentation was really good in most aspects. I was there and thats where I found about locust. Thumbs up!
>
> Though, it was really not mapserver vs linux, more like IIS vs Apache.
>
> And yes, Alexandre is most correct, so much is about configuration.
>
> Quick share of my test about 3 years ago. Do not have original data but some presentation had snippets of it.
> Requests are taken from real access logs. Always the same combo of different extents and layers. Flushing cashes and restarting systems between tests etc...
>
> Apache (original setup without much tuning) vs nginx with unix sockets and some other tunings.
>
> Avg (46 requests per second):
> Apache: 147ms
> Nginx: 96ms
>
> Avg (73 requests per second):
> A: 1400ms
> N: 102ms
>
> In the end, use what you are comfortable with. Remember that every change must be tested and tested again.
>
>
> Cheers!
> Märt
>
>
> -----Original Message-----
> From: mapserver-users [mailto:mapserver-users-bounces at lists.osgeo.org] On Behalf Of Jeff McKenna
> Sent: Wednesday, October 05, 2016 7:01 PM
> To: mapserver-users at lists.osgeo.org
> Subject: Re: [mapserver-users] Mapserver speed/load test
>
> On 2016-10-05 12:45 PM, Leroux, Alexandre (EC) wrote:
>> Hi all,
>>
>> On this topic, I remember seeing this FOSS4G 2016 presentation named
>> 2016 Stress Testing and Analysing MapServer Performance
>> <http://ftp5.gwdg.de/pub/misc/openstreetmap/FOSS4G-2016/foss4g-2016-11
>> 07-stress_testing_and_analysing_mapserver_performance-hd.mp4>
>> (http://ftp5.gwdg.de/pub/misc/openstreetmap/FOSS4G-2016/foss4g-2016-11
>> 07-stress_testing_and_analysing_mapserver_performance-hd.mp4
>> ). Surprisingly, in this speaker's experience, MapServer has been much
>> faster on Windows than on Linux, see his talk / slides that begin
>> slightly after 15:20. I'm really no expert on the topic, but maybe
>> this info can serve in this discussion (of course, configuration/fine
>> tuning is probably crucial to performances)?
>>
> Yes thanks for mentioning that presentation Alex.  I also confirmed with Seth (the presenter) that they run MS4W on their public-facing servers.
>
> Thank you to Seth for sharing those results!
>
> -jeff
>
>
> --
> Jeff McKenna
> Developers of MS4W
> http://www.gatewaygeomatics.com/
>
>
>
>
>
>
> _______________________________________________
> mapserver-users mailing list
> mapserver-users at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-users
> _______________________________________________
> mapserver-users mailing list
> mapserver-users at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-users


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/mapserver-users/attachments/20161005/97a58321/attachment.html>