I tend to make new databases based on function. If we get a lot of data from the IDMS system of the mainframe, then it's likely to wind up in a database called idms. If we have a database for the company phonebook, it winds up in a database called phonebook.
The important thing to remember is that it is hard / non-standard / poor for performance to do things across multiple databases. i.e. if you have related data that need to be joined up together, it should reside in the same database.
If the data aren't related, then it's likely that you would be better served putting them into seperate database instances for performance and security reasons.
