Questions To Address

Here are some questions that you should address before implementing a database:

Will performance be acceptable for the five to ten most important transactions? From a performance standpoint, the most important transactions are those transactions that are executed most frequently.
Do any clustered entities require rapid, random retrieval? If so, consider placing indexes on these entities or, in a non-SQL implementation, adding additional linked relationships, as described below.
In the following example, the EMPLOYEE entity is stored clustered via the DEPT-EMPLOYEE relationship. A new entity called EMP-NUM is created and linked to the EMPLOYEE entity in a one-to-one relationship. Using the relationship and CALC retrieval on EMP-NUM, an employee can be retrieved by employee number using two I/Os, even though it is neither a CALC nor an index key.

Does any entity that sparsely populates an area require processing of all occurrences? If so, consider building an index for the entity.
Can extra relationships be added for more direct access? In some cases, you might want to include additional relationships to enhance processing performance. For example, you might want to define the DEPT-SKILL relationship to allow retrieval of information from the DEPARTMENT and SKILL entities without having to retrieve employees. The diagram below shows this use of an extra relationship.