Experimenting with data to determine value - migration/methods?

I have a LOT of data available to me, and want to capture and experiment with data that isn'开发者_如何学Pythont currently being used in production. I do not want to immediately add this to my existing data store since that would undoubtedly mess with production. The obvious solution seems to be to make a copy of production data and integrate it with what I want to play around with (applications accessing this data ,etc), but I was wondering if there was a better (less expensive?) way to do this.

Both isolation and integration are important. I'd like to be able to keep lightweight/experimental data assets apart from high volume production data, but also be able to integrate (RELATIVELY) painlessly if experimental assets are deemed useful.

Thanks.

I would prefer to work on small amount of data initially to start with and create a duplicate of the production environment and then start working on the migration part.

If i am successfull, within a less time, then the same steps would be continued for the rest.