GMelanie

It's always better to find out what the problem is, most likely your "script" whatever
that software is, is not letting go of native OS resources, like database handles, file handles etc.. or is doing something inefficient with SQL or files etc.. It is found that if you have to throw hardware at the problem, you delay the inevitable, throwing more hardware at the problem (unless you have a very special case).

You don't, you make a shell account on the box and use standard profiling tools, top, sar, ps, uptime come to mind at first. Non withstanding code profiling tools to see speed of execution etc, per DB call, per script call, per page render.

Less than %65 percent of all total, above this number the OS starts to struggle with opening new connections, swapping and opening files, so the effect is compound at about an x(squared) after that, where each subsequent request eats more and more until peak %100 is unresponsive.