-- Benjamin Eberlei <[EMAIL PROTECTED]> wrote (on Thursday, 28 August 2008, 11:06 AM +0200): > Rasmus Lerdorf had a new talk (froscon08) on framework performance in general > and vs using no framework at all and came up with ZF being roughly in the > mid-field (Cake being way slower, CodeIgniter being faster). > > He also talked about some optimizing strategies regarding include path, and > the include strategy in general, so i took some time to optimize within the > Zend Framework and found interesting results: Stripping all require_once from > the complete Zend Framework code, and require (without once) all file > dependancies directly in your main script gives you a performance boost of > about 20-40% for each request.
This is actually something that a lot of performance experts recommend already, and something I want to approach with 2.0. Other projects that are doing autoloading or discussing it are also adding some logging to see what files are loaded per request -- which allows you to generate the list of requires to use as you state above. When you use this strategy alongside an opcode cache, even if not all classes are necessary for each given request, you get a pretty significant boost. The one place where this will not work is plugins, particularly if you have project-specific plugins that override functionality, but which are only used in some areas of the site. However, I think we can likely figure out ways around this as well. How did you strip the require statements? Did you use a script? inclued? or...? > See all the different include strategies and their numbers here: > > http://www.whitewashing.de/blog/articles/73 > > If requiring all the dependant files up front is to hard to find out, you > could still optimize performance of the Zend Framework by fixing your include > path (put /usr/share/php in front of the dot, rather the default config which > work the other way round) and strip all require_once 'Zend/*'; code from > your downloaded ZF library source code. > > > On Tuesday 26 August 2008 19:27:27 Endijs Lisovskis wrote: > > At first let me say THANKS for your reply! I was not expecting such a long > > comment. I will definitely look at ZF 1.7 features and optimizations. > > > > I can't give you any links to articles where someone is pointing out that > > ZF is slow, except one you already gave. But each time when I talk about > > frameworks and ZF in particular I receive negative reactions and all of > > theme are because of doubts about ZF performance. Those comments I receive > > in my blog where I talk about ZF and even maybe promote it and in local PHP > > forum too. > > OK - those who comment cant provide exact numbers in which I could see > > difference between ZF and other frameworks. But all they need to say - ZF > > is complicated, big and with a lot of components etc. etc. And each version > > is becoming bigger and bigger. I would like to argue, that growth in > > features is not affecting speed - but I can't because there are no any > > tests out there. > > I hope you see the problem. In short "They say it is slow, but there is no > > way we can prove they are wrong". > > > > And sorry for my grammar and structure of sentences. English is not my > > mothers tongue, so it is not very easy for me to write in correct English. > > > > wllm wrote: > > > Where have you heard that ZF is one of the slowest frameworks out there? > > > While I haven't heard many claims that ZF is the most performant- that > > > distinction is usually reserved for Code Igniter or one of the other > > > lightweight, performance-focused frameworks- I haven't heard any claims > > > that it is the least performant. The most transparent and sound > > > benchmarks that I've seen are available here: > > > http://www.avnetlabs.com/php/php-framework-comparison-benchmarks. That > > > puts ZF at roughly 50% of the throughput of Code Igniter. That may sound > > > like a lot at first, but as I mention in a comment there, baseline PHP > > > can handle *15 times* the load than Code Igniter and CakePHP falls far > > > short of both (at least in these particular tests). This article > > > demonstrates well that accurate performance characterizations among > > > different frameworks are *extremely* difficult to come up with in the > > > first place, and that any framework faces tradeoffs between performance > > > and functionality. In the end, I'm pretty comfortable with the > > > characterization that ZF is 'twice as slow' as a framework that has been > > > optimized from the get go for performance like Code Igniter. Those who > > > understand the tests at hand and have real performance requirements will > > > understand that this may easily be eclipsed by functionality > > > requirements for any given project. > > > That said, we are planning to focus all of the Zend team's contributions > > > for 1.7 around performance analysis and enhancements. We don't have a > > > test environment with which we can thoroughly benchmark all the > > > frameworks yet, but we will have one set up soon. Once we can start > > > getting good numbers, we will focus on the performance of the MVC > > > components, along with other components- like Zend_Search_Lucene- which > > > are very sensitive to algorithmic complexity and optimizations. We've > > > already profiled the framework with a simple application and identified > > > some optimizations. If you are interested in the performance of ZF, then > > > keep your eyes peeled for optimization commits in the next few weeks. > > > > > > ,Wil > > > > > >> -----Original Message----- > > >> From: Endijs Lisovskis [mailto:[EMAIL PROTECTED] > > >> Sent: Tuesday, August 26, 2008 8:30 AM > > >> To: [email protected] > > >> Subject: [fw-general] Speed and performance between ZF releases > > >> > > >> > > >> Hi! > > >> > > >> I wanted to ask - are there any tests done to compare ZF releases to > > >> see > > >> which ones are faster and uses less resources? > > >> > > >> I'm asking this, because when there are discussions about frameworks - > > >> almost everyone says that ZF is one of slowest frameworks out there > > > > > > (if > > > > > >> compared to Symfony, Cake etc.). It would be good to know if ZF is > > >> making > > >> progress, or failing because of all new functions added to it. > > >> > > >> Thanks! > > >> > > >> Endijs Lisovskis > > >> -- > > >> View this message in context: http://www.nabble.com/Speed-and- > > >> performance-between-ZF-releases-tp19164298p19164298.html > > >> Sent from the Zend Framework mailing list archive at Nabble.com. > > > > -- > Benjamin Eberlei > http://www.beberlei.de > -- Matthew Weier O'Phinney Software Architect | [EMAIL PROTECTED] Zend Framework | http://framework.zend.com/
