Document Type

Dissertation

Publication Date

1995

Abstract

Many data management environments face a critical need to integrate heterogeneous data-data that are stored in varying locations using various data management systems with diverse data formats and schemas. To address this problem, the database research community has developed the concept of a heterogeneous database system (HDB) that provides users with the illusion of a single unified database. However, HDBs rely on the implicit assumption that all data to be integrated into the HDB are stored in full-fledged database management systems (DBMS). This assumption leaves environments that need to integrate non-DBMS data unserved by HDB systems. Furthermore, HDBs are complex software solutions that are not easily lmplementable by database developers wrestling with heterogeneous data. This thesis presents a new, easily implemented HDB architecture that is suitable for integrating non-DBMS data.

The key to our architecture is using an object-oriented database management system (OODBMS) as an implementation tool. Rather than developing an HDB from scratch, we leverage the power and facilities of the underlying OODBMS to provide a query language, application programmer interface, interactive query interface, concurrency control, etc. Using object-oriented technology gives us an additional benefit-our HDB becomes an object-oriented HDB (OOHDB) providing users with greater data model expressivity along with a powerful behavioral component.

The OOHDB architecture we present is independent of a particular OODBMS and can be implemented using a number of commercial OODBMSs for a variety of data management environments. We describe one implementation of our architecture using the GemStone OODBMS for accessing heterogeneous materials science data. This implementation demonstrates how easily the architecture can be implemented. We use this implementation to analyze the performance of the architecture and examine the effectiveness of strategies for enhancing performance.

We conclude that for many environments with heterogeneous non-DBMS data, our OOHDB architecture provides a good solution that is easy to implement using commercial OODBMS technology.

Comments

Chair committee includes...

David Meier, Professor
Thesis Research Adviser

Lois Delcambre
Associate Professor

Jonathan Walpole, Associate Professor

James Stanley, Assistant Professor
Materials Science Department

Download

Share

COinS