Greedy file-a new data organization concept for partial match retrieval

Abstract

A method is presented for arranging a set of records into buckets to facilitate partial match retrieval. The idea is mainly motivated by the concept of optimal Cartesian product files. For a set of records, the method recursively partitions them into a collection of optimal Cartesian product files. A performance formula of this new file structure is derived and the superiority of this data organization concept is established. It is shown that the Cartesian product file is only a special case of the proposed file system.<>