Object-Oriented Database Integration

M. Schrefl
PhD Thesis
PTE8801 (1988)
Dissertation, ausgeführt an der Technische Universität Wien, June 1988.

Abstract (English)

Database integration refers to the production of a homogenous interface to a set of autonomous database systems. The problem is to overcome differences between data formats, differences between heterogeneous data models, and differences between semantically inhomogeneous representations of identical of similar facts.

The thesis proposes a three-phase methodology to solve the problem: In the first phase the local database administrators extend the existing databases by object-oriented interfaces. This is done to provide data model homogeneous descriptions of the different database schemata which are semantic oriented and hide the local data representations through the principle of data abstraction. For this purpose basic concepts of an object-oriented data model are defined. The process for providing an object-oriented interface to an existing data base schema is not topic of the thesis.

In the second phase a multi-database view is developed. The principle of generalization has been used to integrate entity types of different non-object-oriented databases. It is shown how the process of object class definition by generalization can be incorporated into object-oriented systems. Different types of generalization classes as well as upward inheritance help the designer to keep the integration effort small.

The principle of message forwarding is introduced to overcome semantically inhomogeneous representations in the databases to be integrated: If an object can not answer a message directly because of structural differences, a message forwarding plan will be defined. The plan will state how the message is to be forwarded to several related objects which can handle the message and how their responses are to be combined.

In the third phase the databases are accessed by the end user through the multi-database view. Since it seems unfeasible to develop a complete global view at any time, especially if the involved systems are large or change rather frequently, it will not be required that an object class in the multidatabase view has to be defined fully before it can be accessed by an end user. A dynamic knowledge acquisition approach is proposed to complete a global user view if needed: If an object can not answer a message directly or by an already defined message forwarding plan it will initiate an inquiry message, whose purpose is to discover dynamically possible paths of a message forwarding plan, which answers the query.