I think the main point is not having to store a duplicate of the 600GB of input data.