This research develops a novel database system architecture that efficiently supports queries over rich data types like multi-media, time-series, matrices, and documents. Complex data types are modeled as Enhanced Abstract Data Types (E-ADTs), which are software modules with a common interface. While the uniform interface leads to extensibility, it also exposes semantic information that can be used for query optimization. The E-ADT architecture is demonstrated through the development of the PREDATOR database system. The project studies optimizations that are essential to efficiently support each data type, and extensions of these techniques across multiple data types. While initial results demonstrate major performance improvements compared to existing database systems, several issues are being explored further. What should be the contents of a toolkit for data type development? Can the mechanism for extensible types be extended to implement a heterogeneous database system? How does one build an extensible database system in a Web environment? The project is developing solutions to all these issues and demonstrating them in PREDATOR. The educational goal of the project is to provide a freely available, full-function, modern database system for the database community to use for educational and research purposes. To this end, the code base of PREDATOR is being released, and the functionality of the system is being augmented with many standard database features. This project has the potential to change the fundamental architecture of the next generation of object-relational systems, with applications ranging from geographic information systems to multimedia databases.
