Parallel processing of datasets is not always useful. The structure of the file which the interface program is reading will dictate whether this facility may be usefully employed. A case where parallel processing is valid would be a results file where more than one results type is stored on the same record for each element. Here, the use of parallel processing would enable both sets of results to be processed in a single pass of the results file rather than two passes. This is the only type of application where parallel processing of datasets will give any speed benefit over sequential processing. The use of parallel datasets carries a small overhead in internal processing, and therefore should only be used where relevant.