When Did Performance Stop Being Important?
Now that I'm finally starting to "get" the Zend Framework I'm starting to have some serious doubts on whether I made the right choice; not in choosing Zend over another framework but in choosing any framework at all. The memory usage is just abysmal across the board and after working with the Zend Framework for about a month or so it's not entirely clear if it's going to scale as I need it to.
Which lead to the question of why. At the moment it seems like a question of speed of development versus performance (which is ironic because Zend Framework is not easy or speedy to develop with).
To be fair, it's not just frameworks that have an uncomfortable overhead. Just take a look at Joomla and Drupal; 2 popular content management systems with an absurd overhead. It's just easier to focus on my current interest rather than the CMS'es.
One thing I'm having a hard time getting comfortable with is how much memory is required when using a php web framework. Out of the box both Zend and Symfony (for example) use around 5MB per request. Understand, this is without any custom code. Just setting up the MVC and Autoloader for the default views and models. Nothing impressive or useful and 5 fucking MB to run that?
After having been on the wrong end of this issue on my own code I'm pretty sensitive to how my code performs; I've written some nasty algorithms and watching them crumble in real time has a tendency to turn you around
Researching the issue doesn't really help. There's a lot of advice on how to improve the performance but it seems to always center around common sense improvements you should be using anyway.
The most touted improvement I've heard is that you have to use a PHP accelerator and opcode cache. I just find that response flawed but not because it's bad advice but because it's common sense. Yes, it's true, but not using a framework in combination with a PHP accelerator and opcode cache is still better in my experience. All relying on those tools does is move the baseline for performance, which you're supposed to do already, and a framework still consumes a good amount of resources on it's own.
In my experience you get about a 50% reduction in memory usage when using something like x-cache but using the Zend Framework still leaves a total of 2.5MB of memory usage to accomplish the bare minimum setup.
One saving grace is that hardware is cheap. Scaling with hardware is usually the go-to escape when the bottleneck is the code but it's not without it's own set of issues. For one thing while it's true that hardware is cheap the labor to maintain that hardware is not. Especially if you want to maintain the server in a proper and responsible manner.
Another option, that's really only available when using the Zend Framework, is decoupling the project from a direct dependence and not use the MVC components. In anticipation of doing this I've been writing a lot my recent code and projects in a style that'll allow easy(ish) separation when the time comes.
At this point I haven't used a framework in a production environment so all of this consternation might be for nothing. I just have a hard time accepting the performance hit of half a MB for using something trivial like a content management system (drupal) or, for example, a component like Zend_Navigation compared to the benefit. What are they actually doing to make the cost worth while?
Still another option is to just walk away from this whole OOP thing and head back to the familiar touch of procedural php and using functions and classes as more of decorators to apply than core components. From my personal experience, and only my experience, using OOP is way more expensive than procedural. At the end of the day I need my programs to work fast, be easy to operate for my users and have a low impact on the server. How does using OOP help that?
At the moment I'm not sure how this is going to work out. I am confident it'll be an adventure though. Hopefully, I find out how Zend will scale before a project of mine goes viral or gets popular. Hopefully.