trade off performance / node number

Pietro_La_Torre

I'd like to know which is the trade off between performance and node number.

Suppose I have a big fact table and lots of dimensions, which would be the best number of node in my cluster? what is the best way to determine node number?

thanks,
Pietro