after week of learning and experimenting with vary solutions of scaling database we choose to use slony-i to replicate very large and loaded tables. there were a few changes in our application, we just added one more connection (persistent unit) to EJB beans for slave replicated database.
solution was really fast to apply and inexpensive.