Dataflows and large datasets

Thanks v-xicai but setting the incremental load in itself is not the issue.  I have done this several times and I'm happy with it, and also with connecting dataflows to other sources etc.

The problem is figuring out the best way to work with existing datasets based on either source data (in databases) or from dataflows, that have quite a lot of history.

Please see my reply to Jeff for additional information.

The single question that I would like to be answered is:

"What is the optimal way to make changes to a large dataset, with signficant historical data, that has a minimal impact on my organisation's users?"

Thanks