Ther are some specialist forums dealing with Capacity Planning and Capacity Management topics. Google should help you find these.

The main thing you need to focus on at the start is your methodology, not on the maths. The above forums should help with the methodology.

From what I know about CP, you need to model what resources an actual instance will take and extrapolate from there. If you have anything running at present that is a useable model then use it, otherwise you will have to build a model. Anything else is guesswork.

When you have your model, look at how much CPU, I/O per second, disk space, etc is being used for a single instance. Then multiply that out to give what you need for the whole application.

You then need to slice that up by what your equipment can handle. If you need 1m I/O per second, do you need multiple HBAs into your SAN to achieve this, or do you even need multiple SANS. Can yuor favourite server handle the CPU and I/O load, or do you need multiple servers. Do the same with every part of the infrastructure. Eventually you will get a shopping list of what to buy.

One of the top CP experts in the UK told me some time ago that the best you will get with this planning is +/- 25% of reality. For a newcomer to this work, +/- 50% would be a good result. Make sure your management have the right expectations about the result of this exercise. If they do not like your figures, suggest they either pay for experienced consultants or accept the best you can offer.