I'm experimenting with Oracle Multi-Master Replication (MMR).  This was the most powerful type of database replication that Oracle offered for 8i (my current production database) and it allows you, with some exceptions, to replicate the database in a higher release (say, 9i) and  pick and choose the objects you want to replicate.  I plan on using it to replicate my current production database to my new 9i database server, the beauty of it being the new server's data will be in sync with the current production data, making cutover immediate ( I almost used the word 'seamless').

With multi-master replication, you can replicate more than just database tables.  You can also replicate indexes, procedures, packages, functions, triggers and user-defined types.  It also updates or modifies the objects being replicated.  For example, adding a column to a table will be replicated on the other server(s). 

Of  course, there are some downsides.  Replicating large databases can require large network bandwith (not only does multi-master push and pull changes between servers, it also sends acknowledgements and administrative data).  MMR uses triggers and procedures, and when a lot of data being replicated, this can result in a substantial performance hit.  Creating and monitoring the replication environment can get complicated. 

I'm incrementally increasing the number of objects in my multi-master test, with the goal of minimizing downtime when switching to the new server.