I tried to use the chunksize and couldn't figure it out (using pandas.SQLTableDataSet). I would be interested in seeing a hello world code snippet once you get it to work.