Query pre-execution and batching in Paradise: a two-pronged approach to the efficient processing of queries on tape-resident data sets