What partitioning method would yield the most even distribution of the data without duplication?