Entity identity, performance, and storage

Abstract

When a database designer needs to model an entity that exists in a domain that does not have a global name authority, the designer must resort to using surrogate identifiers. Traditionally, these entities were referred to as Weak-Entities in Entity-Relationship models. There are many choices available to a database modeler, but each choice comes with sacrifices. This paper documents an experiment that evaluates several of these choices and hypothesizes the best solution from both a performance and a storage perspective. We then apply the suggested best algorithm to a production system and discover tremendous savings in disk space requirements and execution time on certain queries.