what you want in case of the table that gets read a lot, is partition + cluster.
For table that changes a lot, I don't really know myself either. It is usually trial and error type of scenario for me. (partitioning too, but that depends on what parts are updated and under what conditions, etc).
