Well, for a start, you should run benchmarks against an otherwise idle test machine. This machine will of course, be doing NOTHING else, so you can accurately measure the resources required by PHP alone.
Disc I/O is a tricky one, as you really need to see the big picture - if you're using an external database (e.g. anything except sqlite) most of the I/O will probably be done by the database process(es). You can measure this reasonably accurately if you have a separate database server, but not easily otherwise.
The amount of IO required will of course vary depending on the amount of available memory (due to caching) - so you will need test boxes with the same RAM as the production machines.
I can forcast that in all likelihood, on a web application server running PHP as its only server-side code (and no local database processes or other major services), PHP itself will be consuming all most all of the processing time - web server time for other tasks will probably be negligible.
If you need to monitor individual requests, either configure apache to time them (you can do this in microseconds using the standard Apache logging mechanism) or use a PHP profiler (e.g. APD) or ad-hoc profiling kit (e.g. call microtime a lot)
Mark