Thread: Interesting Benchmark Article

Stumbled across this post by Joe Stagner of Microsoft:
http://misfitgeek.com/blog/aspnet/php-linux-windows-asp-net-performance-ndash-redux/

It is basically a comparisson of ASP.NET vs PHP but it includes
comparissons of multiple RDBMS and operating systems.

Postgres comes out on top for most of the benchmarks against MySQL and
SQL Server, in fact in the authors original article he goes as far as
recommending using Postgres. More interestingly, the article seems to
indicate that W2K8 server is faster for postgres than Linux in many of
the scenarios.

Note that the linux version they were comparing was Ubuntu - which I
know from personal experience has very poor default settings for
shared_buffers due to Ubuntu's operating system defaults. Perhaps if
theses parameters were altered postgres would get a clean sweep.

Anyway - I know benchmarks of this type tend to be very limited in their
scope but good press all the same!

Howard
www.selestial.com

Howard Cole wrote:
> Postgres comes out on top for most of the benchmarks against MySQL and
> SQL Server, in fact in the authors original article he goes as far as
> recommending using Postgres. More interestingly, the article seems to
> indicate that W2K8 server is faster for postgres than Linux in many of
> the scenarios.
I wouldn't go that far.  The numbers are all really close for the
PostgreSQL results, and his results don't deserve nearly as many
significant digits as shown on his spreadsheet.  And the PHP
implementation used has more impact on results in some cases than the OS
change.  For all we know the entirety of the difference relates to, say,
the quality of the PostgreSQL PHP driver on the two platforms.  There's
not enough data here to prove anything beyond "for this simple PHP test,
PostgreSQL performs the same on Windows and Linux, MySQL matches the
Linux results, and MySQL lags far behind any of those on Win Server
2008."  Which is, again, not necessarily a statement about MySQL
performance vs. PostgreSQL performance at all--for all we know it's
because the MySQL PHP driver is garbage.

> Note that the linux version they were comparing was Ubuntu - which I
> know from personal experience has very poor default settings for
> shared_buffers due to Ubuntu's operating system defaults. Perhaps if
> theses parameters were altered postgres would get a clean sweep.
The parameters are no better on a default Windows install.

--
Greg Smith    2ndQuadrant   Baltimore, MD
PostgreSQL Training, Services and Support
greg@2ndQuadrant.com  www.2ndQuadrant.com

Greg Smith wrote:
> Howard Cole wrote:
>> Postgres comes out on top for most of the benchmarks against MySQL
>> and SQL Server, in fact in the authors original article he goes as
>> far as recommending using Postgres. More interestingly, the article
>> seems to indicate that W2K8 server is faster for postgres than Linux
>> in many of the scenarios.
> I wouldn't go that far.  The numbers are all really close for the
> PostgreSQL results, and his results don't deserve nearly as many
> significant digits as shown on his spreadsheet.  And the PHP
> implementation used has more impact on results in some cases than the
> OS change.  For all we know the entirety of the difference relates to,
> say, the quality of the PostgreSQL PHP driver on the two platforms.
> There's not enough data here to prove anything beyond "for this simple
> PHP test, PostgreSQL performs the same on Windows and Linux, MySQL
> matches the Linux results, and MySQL lags far behind any of those on
> Win Server 2008."  Which is, again, not necessarily a statement about
> MySQL performance vs. PostgreSQL performance at all--for all we know
> it's because the MySQL PHP driver is garbage.
>
Hey Greg, I did not write the benchmarks - the benchmarks are comparing
ASP.NET with PHP  and if I was using either then you would want to take
into account the drivers because they are all part of the same package
as far as developers care. Benchmarks are one-off comparissons and I do
not claim anything about the ones posted in the article other than they
are good press for Postgres -especially coming from a Microsoft employee
- especially since they provide some elusive benchmarks on SQL Server.

>> Note that the linux version they were comparing was Ubuntu - which I
>> know from personal experience has very poor default settings for
>> shared_buffers due to Ubuntu's operating system defaults. Perhaps if
>> theses parameters were altered postgres would get a clean sweep.
> The parameters are no better on a default Windows install.
>
The postgres wiki notes lead me to believe that the shared_buffers
settings are not as important on windows as Linux. I am not going to run
the tests to find out - I am accepting these results as in the light
they are given - Pro Postgres :)

Howard Cole
www.selestial.com