Higher Order Databases and Multimedia Information

This paper has two parts. In the first part, we give an overview of higher order databases, the research framework and vision of the database group at ETH Zurich. Higher order databases will be or at least will contribute to the infrastructure for distributed information systems engineering of the future. While a traditional database administers data, a higher order database administers components and specialized services. In particular such components can be databases. Then a higher order database is a coordinator of a database cluster. In general, a higher order database coordinates the components of the cluster together with components dedicated to specific tasks. In the second part we apply higher order databases to multimedia information systems. We show how a networked image retrieval and management system can be built. Its components consist of image servers, feature extraction services and specialized indexing components allowing fast image similarity search. They are coordinated by a DBHO through transactional processes that realize similarity search, insertion and bulk insertion of images. The key features of our system are: scalability with respect to feature extraction services, parallelization of indexing components by replication and adaptation to dynamic changes of available feature extraction components. Part I: Higher Order Databases