Foundations for multifile design by application partitioning

Two principal characteristics of database systems are the !.arcc volume of data and the diversity of application programs' requirements. Individual application programs, however, typically require a relatively small fraction of the data -often values from a subset of the records and almost certainly values from only a few of the attributes. This observation leads directly to the consideration of clustering techniques for distributed databases or, when geographical distribution is not required, for file segmentation.