A Study on the Effect of a Table’s Involvement in Foreign Keys to its Schema Evolution

In this paper, we study the evolution of tables in a schema with respect to the structure of the foreign keys to which tables are related. We organize a hierarchy of topological complexity for the structure of foreign keys, based on a modeling of schemata as graphs, where tables are classified in increasing order of complexity as: isolated (not involved in foreign keys), source (with outgoing foreign keys only), lookup (with incoming foreign keys only) and internal (with both kinds). Our study reveals that this hierarchy reflects also the update behavior of tables: topologically simple tables are more likely to have a life with few or zero schema updates, whereas, topologically complex tables are more likely to undergo high numbers of updates. Early versions of the database attract the large majority of births of complex tables, as opposed to the simple ones, demonstrating a pattern of reducing the introduction of complex, heavily updated constructs in the schema as time progresses.