Economic incentives for database normalization

It is suggested that normalization reduces the opportunity cost associated with information retrieval from a database by improving recall and is most pronounced when recall is low, which can create fragmentation inefficiencies and information overload.

Abstract

Database systems are central to business information processing. The conceptual basis for most commercial database managers is the relational model. Little research exists concerning the cost effectiveness of relational database normalization, but there is anecdotal evidence that normalization-induced fragmentation may create inefficiencies. Supply and demand for normalization is investigated given management policies for response time, database capacity, and deletion policies. On the supply side, normalization reduces costs associated with insertion, deletion, and change anomalies. The expected cost of removing change anomalies is linearly proportional to both minimum database size and to database capacity. The occurrence rates of either insertion or deletion anomalies are shown to be moderate for all but microcomputer sized databases. But because insertion or deletion anomalies tend to result in significant cost, even small probabilities of occurrence can result in significant costs. On the demand side, normalization can create retrieval inefficiencies where a comparatively small amount of information is being sought and retrieved from the database. Both an increase in clustering, and an increase in database size will exacerbate these inefficiencies. This can result in fragmentation inefficiencies and information overload. It is suggested that normalization reduces the opportunity cost associated with information retrieval from a database by improving recall and is most pronounced when recall is low. Where retrieval rates are high with respect to update events, the database fragmentation caused by normalization costs end users through slower retrieval response.