Building a Federation of Heterogeneous (Non-Database) Repositories
|Title||Building a Federation of Heterogeneous (Non-Database) Repositories|
|Author(s)||M. Tresch, U. Röhm|
|Booktitle||W. Hasselbring (Hrsg.), Proceedings of the 2nd Workshop Föderierte Datenbanken, pages 57-63Dortmund, Germany|
|Organization||SWT Memo Nr. 90, Fachbereich Informatik, Universität Dortmund, 1996|
AbstractWe are developing a federated database management system that provides (i) a uniform view of and (ii) advanced database functionality over data externally stored in repositories. Our particular interest are along the following two main lines of investigation:
Service-Oriented Federation. We propose a service-oriented federation layer, that provides selected database services, like for example data modeling, integrity constraints, query processing/optimization or transaction mechanisms, for externally stored data. The goal is to achieve a federation layer that is as thin as possible and as thick as necessary. Current approaches to federated databases are extended in that we allow component systems to be non-database management systems. In particular, we include repositories like file systems or "black-box" application systems, e.g. a library information system. We aim to export lacking database functionality to such repositories.
Gradual System Integration. While integrating local repositories into a global federated system, only selected parts of the component systems are usually required. Furthermore, many repositories do not provide meta-data that is required to integrate their data.
As a consequence, (semi-) automatic integration methods have in most cases shown to be impractical. We therefore propose a different approach to facilitate the integration of heterogeneous repositories into the federation. So called "object-likeness" (see also the likeness concept of [BRS:SIGMOD96]) is used to gradually integrate repositories and hence stepwise achieve more and more refined object-oriented views of external data.
In this short paper, we focus on providing distributed object management functionality for external non-database repositories and show how build a federated system that exports data modeling, query processing/optimization, and integrity constraints services.