Also, to make sure that nobody misunderstands my intention here; these numbers were tested on two un-optimized installations. Rather than showing that Lighttpd or Apache is faster than the other, my intention is to show that from the PHP standpoint they are close enough in terms of performance that performance concerns are relatively minimal. There will likely be much, much lower hanging fruit than any difference between Lighttpd and Apache would show.

Thanks. The code tracing feature sounds quite interesting.

… also, you don’t NEED FastCGI to run PHP, but if you want to run it in any way that resembles useful, from a performance perspective, you would want to.

Yeah, the “lightweight” refers to the connection architecture. After the connection is made, the PHP is identical with regards to execution, except for the SAPI.

Zend Server is our application server for PHP. You can start at http://www.zend.com/en/products/server/. Generally we try to hook in with whatever server is set up on your machine. IIS for Windows and Apache for Linux/i5. If you have a web server from a supported setup (i.e. RHEL) we will integrate with the installed web server instead of installing our own. We also have a public beta for Zend Server 5.0 with a new feature called Code Tracing which is worth the time to look at. Please let me know if you have any questions.

That’s pretty interesting comparison. The reasoning behind Lighty being faster than Apache is because well it’s lightweight apparently. I wasn’t aware in order to run php, it it utilizes fastCGI. I’m assuming same holds true for Perl.

I have to check out the Zend Server pkg…I’m guessing that’s a commercial package provided and maintained by Zend which contains apache/php/mysql, does regular updates etc? Wouldn’t mind posting an article about it here.

Thanks for your info.

Did some quick testing and saw that they were virtually identical. I used FCGI/PHP/APC for Lighttpd and Zend Server/Apache with a “Hello World” Zend Framework app. Zend Server w/ Apache was 3.9ms per request, Lighttpd w/ FCGI was 4.1ms per request. So, Apache was a little quicker, by 1/5th of a ms.

I doubt it. Once PHP is running it is generally staying inside of the PHP binary and the web server architecture doesn’t matter as much at that point. Lighttpd generally runs PHP from within FastCGI which is a layer in between the web server and PHP. While I don’t have any specific numbers I would actually expect Lighttpd to run PHP a smidgen slower because of the FastCGI layer. Actual numbers would be beneficial so I will see if I can dig some up.

What is often done is running PHP inside of Apache, but then taking static content such as images, CSS and JavaScript, and serving those from Lighttpd. What this does is make your website, as a whole, _feel_ faster. That is generally going to be more important than eeking out a few additional ms of performance from PHP.

Generally you won’t be able to find a lot of things about server settings or php.ini. PHP is pretty optimized already and many of the settings that PHP has available are more for turning specific features on and off. Architectural considerations are more likely to have a larger impact than configuration tuning.