When sizing a cache such as Ehcache, back-of-envelope calculations is what’s often done. While it may well give the right answer, the envelope never seems to be kept so when someone comes along needing to modify/tweak the caching behaviour of the application they are faced with an ehcache.xml config file full of various numbers with no indication of why they were chosen.
This is often the case with many config files, but I’ve seen that ehcache.xmls are particularly prone to this (not anything to do with Ehcache, more the people who maintain them).
So I think it’s good to do what we did in a recent project, and document the choices of various settings with comments, and the calculations in a spreadsheet checked in to source control.
Some resources I found useful while calculating cache sizes: