Performance benchmarks pitting FreeBSD against Windows

[Mel]

On Friday 05 December 2008 17:45:37 Chad Perrin wrote:
> On Fri, Dec 05, 2008 at 02:26:50PM +0100, Mel wrote:
> > Well, one can find stories like this of course:
> > http://www.postgis.org/documentation/casestudies/globexplorer/
> >
> > But I'm sure one can find some of the contrary. It does show the value of
> > the benchmark: Is it economically viable to use configuration X vs Y, and
> > performance is only one factor of the descision.
>
> Actually, the only other story that comes immediately to mind of a
> PostgreSQL vs. Oracle comparison is this one:
>
>   http://www.enterprisedb.com/about/news_events/press_releases/06_27_07.do
>
> . . . so, in my experience at least, stories to the contrary are pretty
> hard to find.
>
> Of course, that seems to be more about PostgreSQL vs. Oracle than FreeBSD
> vs. MS Windows.

Point being, that a benchmark should never decide or even help decide to 
change software accross the board as a policy.
You may use it as orientation, but in practice the value of said benchmarks is 
low as they rarely represent real workloads.
The deciding process is migrating one and see what happens, what you have to 
do to migrate (it's rarely "just the os" and takes man hours) and what the 
difference in maintenance and periodic costs is.
Benchmarks are more useful to "see what kind of hardware I'd need to run a 
MySQL server with X simultanious connections on FreeBSD" and even better if 
the tuning and optimizations for the benchmark are documented.

-- 
Mel

Problem with today's modular software: they start with the modules
    and never get to the software part.