That is kind of a broad question. Do you want a two-phase approach where the commit doesn't happen on the server if the client cannot commit?
Is this a two-way replication. If the client updates/deletes/??? does that transaction need to come back to the server?
How often does this need to happen (real time, once a day)?
For a single table it might be easier to write your code using an internal 'queue' table, triggers and a database link.
The main Oracle replication methods are:
1: Streams replication
2: Oracle Advanced Replication
You can also to some things with Materlalized views and/or transportable tablespaces.
