On Tue, Aug 21, 2012 at 9:17 PM, Gabriel Roldan <[email protected]> wrote:
> I found the graph below in the WMS Benchmarking 2011 presentation result,
>> but I cant understand one thing. It is tested on Linux server or Windows or
>> this is the average from both server and only MapServer is divided into 2
>> values? I need to compare MapServer and GeoServer, but If I take the value
>> MapServer Win FastCGI and GeoServer, it is more or less same, but if I
>> compare Mapserver Linux FastCGI and GeoServer, MapServer is much more
>> better.
>>
>>
> In 2011 the GeoServer tests were run only on Linux, due to lack of
> dedicated time. Also note we didn't have time to optimize the SLD's or
> anything else, an ran only a single geoserver instance, whilst if what
> you're looking for is higher throughput, I think the recommended set up is
> two-to-four geoserver instances even per physical server, due to the Java2D
> rasterizer being synchronized. Others may be able to comment better on the
> issue though.
>
Yep, look at slide 67 in this presentation to see how much better GeoServer
can do when you run multiple instances
on a single physical machine behind a load balancer (please note this
benchmark was run on a notebook, not on a server
like the FOSS4G 2011 one, even if the dataset was the same):
http://demo.geo-solutions.it/share/foss4g2011/gs_steroids_sgiannec_foss4g2011.pdf
The above was done using Apache as the load balancer, using something that
is dedicated to load balacing
(HAProxy) would lead to better performance (see for example this blog entry:
http://geo-solutions.blogspot.it/2012/08/real-world-uses-cases-troubleshooting.html
)
This model is somehow similar to the one used for MapServer in the
benchmarks, where they did use 8 instances pinned down with FastCGI.
The thing that often surprises me is when MapServer users trying out
GeoServer tell me "wow, GeoServer is fast" because few people use the
FastCGI configuration, and without it MapServer is significantly slower
than it was in those tests.
It would be possible to make GeoServer scale better using a single
instance, it's a matter of rewriting the Java rasterizer (which, in Java 7,
is something
that can be done, it has been made pluggable), but it would require quite a
bit of dedicated funding, it's a hard block of
software to write that would keep a core developer busy for a month or two
Cheers
Andrea
--
==
Our support, Your Success! Visit http://opensdi.geo-solutions.it for more
information.
==
Ing. Andrea Aime
@geowolf
Technical Lead
GeoSolutions S.A.S.
Via Poggio alle Viti 1187
55054 Massarosa (LU)
Italy
phone: +39 0584 962313
fax: +39 0584 962313
mob: +39 339 8844549
http://www.geo-solutions.it
http://twitter.com/geosolutions_it
-------------------------------------------------------
------------------------------------------------------------------------------
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
