From recent tests conducted with StreetDirectory.com shows that PHP executes a lot faster on Windows Server 2008 now with their built-in CGI support. In fact it runs 130% faster then any other server operating system out there.
The built-in support of FastCGI enables PHP applications to run efficiently on Windows Server 2008, with significant performance boosts. As can be seen in the graph from the case study:

The Windows Server 2008 installation was a core installation with no performance tweaks, making it a out-of-the-box solution for all server administrators.
As you can note with the graph above that the 4000 users point, Windows Server 2008 becomes slower then Linux. Dennis from Microsoft says many server administrators would evaluate the performance by looking at where 75% of the maximum hits is located.
If we take the same approach from Dennis here, then most server administrators would be scaling out to another machine at about 45,000 sustained hits, whereas on the Linux box it would be around 30,000 sustained hits.
So what does this mean? Less servers for the administrators to maintain, less hardware to have problems with and less network/software issues to deal with. It will also be cheaper and more effective to scale out when needed or for 'just in case' scenarios.
I haven't been brain washed by Microsoft as a Microsoft Partner but I do seriously believe that Microsoft has made dramatic changes to both Windows Server 2008 and Windows Vista. Look out Linux!
Related Articles
* PHP runs up to 130% faster on Windows Server 2008
* Minty White - PHP runs up to 130% faster on Windows Server 2008
* Windows Server 2008: PHP runs up to 130% faster
* As Windows Server 2008 RTMs, Customers and Partners Adopting with Help of New Tools, Training
21 Responses to “PHP runs up to 130% faster on Windows Server 2008”
Thanks for the discussion here from both sides of the fence, really cleared up the report for myself and many readers that stumbled across this post.
I saw the post by Rich, who I assumed was someone reputable at Microsoft or the company in question - and he said that the linux setup was running CGI.
streetdirectory.com is running apache + BSD at this stage, so I'm interested to see where the linux server was running, and why they took the time to switch to BSD instead of windows. it could be a reverse proxy, but that still doesn't say much for their confidence.
I must apologise for believing blindly that the linux setup was running cgi (who is Rich anyway? :) ) but there are still too many unknowns for this developer to take it seriously. Material like this is for pointed haired middle managers though :P
I don't have any "inside" information on the study, just going by what was stated in the article, which like you said is a piece of marketing material. My interpretation is that the code that was running on linux was moved across on to windows and they preferred the results they received from windows.
not sure if they tested with cgi/fastcgi/mod_php but would guess it would at least have to me fastcgi/mod_php to have the level of performance they are seeing.
I understand this document is a marketing play, so we can't expect it to be totally accurate. The title of the piece is misleading, given that FastCGI is already known to perform a lot better than CGI.
On linux + apache most people run mod_php which is faster than both CGI and FastCGI, in that configuration.
The final point is that people should choose the configuration best suited to their load pattern, there is no _single_ way to do things, much to microsoft's chagrin I'm sure. :)
As it is, the benchmarks are comparing "Fast technique on windows vs. worst practice on other operating systems".
Do Microsoft actually realise that CGI != FastCGI?
I'm actually gobsmacked that Microsoft has openly published this crap.
Lets just say that I'm happy I don't use ANYTHING Microsoft. ahh, life without marketing bs is bliss.
I guess that pretty much sums up the evaluation. That comparison brought on a few chuckles in the office here.
http://blogs.technet.com/jorke/archive/2008/04/14/huge-performance-increases-with-php-on-windows-2008.aspx
The previous Linux System was running Apache. A report on how Apache works on Windows is not given.
The Linux server was running CGI. I hope that answers a few of your questions :)
was the linux server running prethreaded, or preforked? or thread pool? was it running cgi, or fastcgi or as a module?
a lot of unknowns!