Database Reference
In-Depth Information
the global data model. The selection of global data model depends on a number of factors including the
semantic richness of the local databases (Saltor, Castellanos and Garcia-Solaco, 1991; Seth and Larson,
1990) and the global application requirements. Nevertheless, the impact of instance integration on the
global data model has not been well studied so far. In this chapter, we study this impact in the context
of fuzzy relational data model.
In this research, we assume that the schema integration process has been carried out to the extent
that a global schema has been obtained from a collection of existing (local) fuzzy relational databases.
Hence, global users or applications will formulate their queries based on the global schema. Moreover,
export schemas that are compatible with respect to the global schema have been defined upon the local
fuzzy databases. We classify instance integration into three distinct levels according to the extent to
which instance integration is carried out:
•
Level-0: Neither entity identification nor attribute value conflict resolution is performed. Since
no instance integration is involved, the integrated database is defined merely by collecting the
instances from different local databases into relations specified by the global schema.
•
Level-1: Entity identification is performed but not attributes value conflict resolution. Hence, local
database instances which correspond to the same real-world objects are matched and combined in
the global relations. However, the attributes of these matching database instances are not merged.
•
Level-2: (complete integration). Both entity identification and attribute value conflicts are re-
solved. In this case, the local database instances are completely integrated.
Earlier research on database integration indicates that complete integration of instances is the only
ideal solution for database integration. Nevertheless, we argue that there are some reasons advocating
different levels of instance integration.
1. Firstly, it may not be possible to acquire sufficient knowledge to perform complete instance
integration.
2. Secondly, data quality of local databases may be low and it is not worthwhile to perform complete
instance integration.
3. Thirdly, performing instance integration may be costly, especially for the case of virtual database
integration in which instance integration is being performed for every global query. For many or-
ganizations, the benefits of complete instance integration may not outweigh costs associated with
the integration.
4. Lastly, in some cases, the global users or applications may not require complete instance integration.
Apart from level-2 instance integration which represents the complete integration, the levels 0 and
1 impose some constraints upon the global data model:
•
Due to incomplete instance integration, the integrated database is expected to accommodate some
remaining instance level heterogeneities. It is the responsibility of global applications to resolve
remaining instance-level conflicts when the need arises.
•
On the other hand, there exists the possibility that the levels 0 and 1 integrated databases may be
needed to be fully integrated with human involvement combined with additional domain knowl-
