What is this going to look like for document stores, columnar databases, key/value pair and other so-called NoSQL databases where the data (or the map/reduce program) is the model? I'm interested to see how these products and vendors evolve. Informatica was pretty quick to jump on the Hadoop train.