ParentFull threadtadlan·Dask has that, and scikit learn is moving that way also. It even beats spark for out of core work on a single machineView on HN