If you're coming from a Java background, you might be used to writing a large infrastructure of classes that rely on each other to perform a particular task.

Don't try this with PHP! PHP's OOP features work best when your framework is small and efficient. Creating objects in PHP is a rather slow process, and, as such, it should be used conscientiously.

Caching Techniques

Sometimes, it's just not possible to optimize your code beyond a certain point. It might be that your queries are too complicated or that you depend on a slow external resource, such as a web service, over which you have no control.

In these cases, you might want to think about using a caching solution that "saves" the output of an operation and then allows you to access it without performing that operation again.

There are several types of cache; for example, you can save the results of a database query, or even an entire web page. The latter means that you generate your pages normally at predetermined intervals and save them in the cache. When a page is requested by a user, it is actually retrieved from the cache instead of being generated from scratch.

You can find several packages in the PEAR repository that are useful for output caching of various type. Naturally, there are also commercial solutions that perform a similar task, such as the ones provided by Zend.

Bytecode Caches

When PHP runs your scripts, it does so in two steps. First, it parses the script itself, transforming it into a sort of intermediate language referred to as bytecode. Then, it actually interprets the bytecode (which is simpler than PHP itself) and executes it. If your scripts don't change between one execution and the next, the first step could easily be skipped, and only the second step would have to be taken.

This is what "bytecode caches" do. They are usually installed as simple extensions to PHP that act in a completely transparent way, caching the bytecode versions of your script and skipping the parsing step unless it is necessary—either because the script has never been parsed before (and, therefore, can't be in the cache yet) or because the original script has changed and the cache needs refreshing.

A number of commercial and open-source bytecode caches (also called accelerators) are available on the market, such as the one contained in the Zend Performance Suite, or the open-source APC. Most often, they also modify the bytecode so as to optimize it by removing unnecessary instructions.

Bytecode caching should always be the last step in your optimization process because no matter how efficient your code is, it's always going to provide you with the same performance boost. And, as a result, it could trick you into a false sense of security that would prevent you from looking at the other performance optimization techniques available.