I am responsible for a scientific visualization application at my company. The app is 2-tier with a C# fat client talking to Oracle 11g. We are getting ready to deploy the app globally and need to come up with a replication strategy as we have found that performance is unacceptable when the client and database are separated by a WAN. Currently, experimental data are loaded in batches of up to 10G using SQLLDR through the direct path. We would like to set up a staging database that would enable us to examine the data before it goes into production. When the data have been certified, we would like to transfer new records on-demand in bulk to multiple replicate databases globally over the WAN. We would like to be able to roll back or recover from data transfer errors.
I have limited experience with Oracle replication and would appreciate suggestions of products, technologies, or strategies that would meet our needs, which seem different from typical transactional replication.