Physical Database Design
While the process of transforming an ORA-SS schema diagram to an NF ORA-SS schema diagram can lead to databases that have no replicated data which in turn leads to a reduction in anomalies, the addition of references may adversely degrade the performance of queries on the database.
In this chapter, we have addressed how replication can be added back into a schema in a controlled manner in order to improve the performance of queries that are frequently asked. We have reviewed how attributes that seldom change are dealt with in relational databases, and how pairings are maintained automatically in hierarchical IMS databases for answering symmetric queries efficiently.
We have described the kinds of replication that can arise in semistructured databases, and how the replication of relatively stable attributes and relationship types can be added, and how pairings between object classes can be maintained. These concepts have roots in the approaches taken by relational and IMS databases to deal with replication and answering symmetric queries respectively. The costs incurred in the replication of data in semistructured databases include the storage cost, the update cost and the performance cost. Recognizing that some attributes and relationship types are relatively stable gives a more realistic picture of the update cost.
Guidelines for semistructured database design are introduced and demonstrated, from requirements gathering, through conceptual modeling to physical modeling. Finally, we have outlined an algorithm for storing XML documents efficiently and consistently in a traditional database system.
Unable to display preview. Download preview PDF.