Ask this question of each entity that exists neither as a parent and child, nor as only a parent.
An entity that acts as a child but not a parent is not usually used as an entry point into the database. This entity often can be stored clustered around one of its parent entities.
Clustering through a relationship is one of the most effective ways of reducing I/Os when related entity occurrences are retrieved together. Therefore, if applications accessing this entity frequently access related entities, you should generally cluster the child entities through the relationship.
Note: If the size of all clustered entity occurrences is large, the benefit of clustering might be negated because it requires several I/Os to access the entire cluster.
Example
An example of a child entity but not a parent is the EXPERTISE entity. An occurrence of this entity is most frequently accessed through its associated EMPLOYEE entity occurrence. Therefore, it can be stored clustered around the EMP-EXPERTISE relationship.
Copyright © 2014 CA.
All rights reserved.
|
|