The only way we can accurately tune this site is on the live traffic over many days. I don't think it is really possible to tune performance very well with tools that flood a server with traffic. I did run ab a few times, but it is really a waste of time. Here is the link:
ab - Apache HTTP server benchmarking tool - Apache HTTP Server
The trouble is that tuning requires setting many parameters over time, like key buffers, query cache, timeouts, max connections, etc. So many parameters and each parameter is different.
I noticed the other day that someone ran a flooding tool against this site, because we reached maximum connnections, which only happens when a flooding tool is used :-)
Please do not do this. Why? Because it completely ruins the tuning analysis. Flooding is not real traffic and it is not even close to helpful.
In other words, what happens with flooding tools is so different than with real world LAMP.
There is some good reading on the net, for example:
Tuning LAMP systems, Part 3: Tuning your MySQL server
But even these recommendations are only useful as background reference, because each site is different, the hardware is different, the traffic is different, the database queries are different, the JavaScript is different, etc etc.
I used to read all of these, and they are useful; but at the end of the day, LAMP tuning is based on knowing the system details, hardware, software, database structure, traffic pattern, query pattern, load etc. Tools cannot come close to simulating the real traffic (load, queries, timeouts, dropped connections, etc), especially on a site like this.