It is worth reiterating that the coverage of the topic here represents a relatively informal overview. It is stressed that whilst relational databases are extremely popular and widespread, they are not the only type of database currently in fashion. Relational databases are particularly suited to applications where data is structured and, for example, are not necessarily the best tools for mass text retrieval.

 

When relational databases were first developed by Codd & Date in the mid 1970's there was a premium on disk storage space. This partly informed the development process. The process of normalising data results in many small (i.e. records with few fields) tables. This means that there is an overhead for each query processed because if the data is held across multiple tables then the DBMS has to JOIN those tables together to produce the RELATION. Such systems are particularly suited to applications which have high levels of transactions e.g. student record systems and banking systems.

The rapid growth in the power of computing hardware and the corresponding fall in the cost of computing hardware means that other software techniques are being developed which would not have been previously feasible. During the 1990's there has been a rapid and sustained growth in the production of OBJECT ORIENTATED systems. Students wishing to further their knowledge of database design are recommended to explore object orientated database techniques.

 

Throughout this exercise we will be asking what happens if we delete, modify and add data. If we discover anomalies we can suspect that our data design is in need of modification.

 

Applying the rules of first, second and third normalisation takes us through the following stages.