That analogy might imply overkill, thus highlighting the tactical advantages of the SFrame approach in processing a month's worth of 1-10GB daily-generated SQLite files, for instance.
Dasks' out of core dataframes are just a thin wrapper around pandas dataframes (aided by the recent improvement in pandas to release the GIL on a bunch of operations)
Why doesn't Pandas have anything to save the entire workspace to disk (like .RData).
There are all these cool file formats like Castra, HDF5, even the vanilla pickle - but I don't see anything with a one shot save of the workspace (something like Dill)
You haven't refuted anything I said. Internally the dask dataframe operations sit on top of pandas dataframes. All dask does is automatically handle the chunking into in-memory pandas dataframes and interpret dask workflows as a series of pandas operations.
That analogy might imply overkill, thus highlighting the tactical advantages of the SFrame approach in processing a month's worth of 1-10GB daily-generated SQLite files, for instance.