I respectfully disagree. Yes, you may need to denormalize, but it isn't the same thing as having multiple identical tables - I still contend that that's a bad idea.

You can still partition the data over multiple servers to achieve better performance, but there shouldn't really be any performance difference between having a single table (with the proper index) and having multiple tables on a single table- unless you have a system where lock granularity is not at the row-level, as with the proper index the same amount of data needs to be read whether you hit a common table or your client's private table.

However all this drifts away from perl, and so isn't really all that pertinent here...

Michael