Using Meta-Structures in Database Design

Practical experience shows that the design of very large database schemata causes severe problems, and no systematic support is provided. In this paper we address this problem. We define an Entity-Relationship schema algebra, which permits the representation of very large database schemata by algebraic expressions involving smaller schemata. Similar to abstraction mechanisms found in semantic data models the schema constructors can be classified into three groups for building associations and collections of subschemata, and for folding subschemata. Furthermore, based on the analysis of a large number of very large database schemata we identify twelve frequently recurring meta-structures in three categories associated with schema construction, lifespan and context. In combination with the schema algebra the meta-structures permit a component-based approach to database schema design, which can further be formalised by graph-rewriting.