Mapserver on Linux vs Windows

[Tim Bowden]

On Sat, 2006-03-04 at 07:59 -0500, Ed McNierney wrote:
> Tim -
> 
> "neither windows or linux were optimised" 
> 
> But doesn't that largely invalidate the test?  

Yes and no.  This was the first round of testing (done before I joined
the team) designed to point the direction in which our efforts should be
focused.  It was functional as much as performance testing.  Our client
(large govt dept) specified open source where possible & ogc compliance,
hence mapserver and geoserver were selected for testing. Both were
tested on windows and linux.  Geoserver suffered from stability and
accuracy issues under heavy load.  We used jserver to ramp up the load
till it broke.  Mapserver in the end didn't break, just got bogged down
when the load was too heavy.  OGC compliance was poor with both (at
least as far as we were able to do the tests).

We are currently doing a new round of testing focused on ms (geoserver
was dropped when our earlier test results were confirmed) with
everything optimised to identify outstanding issues (of which there are
a few).  We will probably be able to publish a summary of our testing if
our client gives permission.

Tim Bowden

> It's very easy to set up poorly-tuned versions of Windows or Linux.  If you simply install both systems out-of-the-box, you're comparing apples and oranges.  This is especially true if you're comparing Red Hat Enterprise Linux vs. Windows XP Pro, a desktop operating system!  XP Pro is most certainly not tuned out-of-the-box to be a server platform.
> 
> If you have actual test data it would be helpful to publish it so others can try to reproduce it.  But I will continue to insist that there is FAR more performance variation in other aspects of the system than in the simple choice of which operating system is used.
> 
> 	- Ed
> 
> Ed McNierney
> President and Chief Mapmaker
> TopoZone.com / Maps a la carte, Inc.
> 73 Princeton Street, Suite 305
> North Chelmsford, MA  01863
> Phone: +1 (978) 251-4242
> Fax: +1 (978) 251-1396
> ed at topozone.com
> 
> -----Original Message-----
> From: UMN MapServer Users List [mailto:MAPSERVER-USERS at LISTS.UMN.EDU] On Behalf Of tim
> Sent: Friday, March 03, 2006 10:08 PM
> To: MAPSERVER-USERS at LISTS.UMN.EDU
> Subject: Re: [UMN_MAPSERVER-USERS] Mapserver on Linux vs Windows
> 
> On Fri, 2006-03-03 at 22:39 +0100, Sture Dingsøyr wrote:
> > Hi there
> > 
> > We have a server running Mapserver on Linux (Red Hat 8). Is is used to generate maps, mainly from shapefiles. But also some layers from Postgis/WMS. To do this we use PHP and MapScrip...
> > 
> > We are now considering the possibility to port our solution from Linux to Windows, mainly du to the fact that maintaining new versions of Mapserver on Windows are quite easy with the pre-built binary package that exists for Windows (no compiling is needed).
> > 
> > Does anybody have any experience on how a Mapserver solution works on Windows compared to Linux. I am mainly thinking about speed and performance? 
> > 
> > Does Mapserver work faster on Linux? 
> > 
> > Regards Sture
> 
> We have been testing mapserver on both linux (rhel 3.x) and windows (xp
> pro) for performance testing.  I can't remember the exact results (someone else on our team did the actual testing) but on high end hardware (Dell dual xeon 3ghz 4gb ram) ms on linux approached 50% higher load than on windows.  I'll have to check next week just how the test were done but the decision was clearly in favour of linux.  I do know that neither windows or linux were optimised and the feeling was we could extend the linux performance considerably (by doing things like not running X etc).
> 
> HTH,
> Tim Bowden
> --
> Mapforge Geospatial
> Level 3/ 267 St Georges Tce
> Perth 6000
> Western Australia
-- 
Mapforge Geospatial
Level 3/ 267 St Georges Tce
Perth 6000
Western Australia