PREDATOR: an OR-DBMS with enhanced data types

We are witnessing an explosion in the volume and complexity of digital information that people want to access and analyze. Much of this data is “multi-media” like images, video, audio, and documents. Several other kinds of complex and semi-structured data are also important, including geographical objects, chemical and biological structures, mathematical entities like matrices and equations, and financial data like time-series. Databaae systems must efficiently support queries over such richly structured datw otherwise, they will fast become “roadklll on the information super-highway” [DeW95]. PREDATOR is an object-relational DBMS being developed at Cornell University. It uses a novel “Enhanced ADT” (E-ADT) technology to meet this challenge. EADTs add semantics to the complex data types, resulting in dramatically improved performance, while retaining the extensibility of the database system. The demonstration shows the following features of PREDATOR: (a) standard relational query processing capability through SQL queries, (b) extensibility with complex data types (images, audio, video, documents), (c) content-baaed feature extraction, indexing and retrieval using path indexes, (d) extensible optimization of E-ADT expressions by reordering and merging, (e) WWW/Javabased user interface that supports extensible querying and display mechanisms. These capabilities are demonstrated individually, aa well aa in the context of the following applications : (1) a multimedia digital library application involving thousands of images from museums of rare art, (2) a GIS application demonstrating raater, point and polygon data, and multi-dimensional indexing. The PREDATOR system will be publicly awailable in summer 1997, along with source code. The goal of the releaae of this software is for it to be widely used for research and educational purposes. We will provide extensive webbased documentation of both the usage and the internal design details. No other freely available research system provides these advanced features and capabilities; consequently, we hope that PREDATOR will be used aa a test-bed for a variety of research ideaa.

[1]  Michael Stonebraker,et al.  The Implementation of Postgres , 1990, IEEE Trans. Knowl. Data Eng..

[2]  Stephen N. Zilles,et al.  Programming with abstract data types , 1974, SIGPLAN Symposium on Very High Level Languages.

[3]  Miron Livny,et al.  The Design and Implementation of a Sequence Database System , 1996, VLDB.

[4]  Michael Stonebraker,et al.  The SEQUOIA 2000 storage benchmark , 1993, SIGMOD '93.

[5]  David J. DeWitt,et al.  Shoring up persistent applications , 1994, SIGMOD '94.

[6]  Michael Stonebraker,et al.  Application of Abstract Data Types and Abstract Indices to CAD Data Bases , 1986, Engineering Design Applications.

[7]  Hamid Pirahesh,et al.  Extensible/rule based query rewrite optimization in Starburst , 1992, SIGMOD '92.

[8]  David J. DeWitt,et al.  Client-Server Paradise , 1994, VLDB.

[9]  Michael Stonebraker,et al.  Inclusion of new types in relational data base systems , 1986, 1986 IEEE Second International Conference on Data Engineering.

[10]  Miron Livny,et al.  The Case for Enhanced Abstract Data Types , 1997, VLDB.

[11]  Patricia G. Selinger,et al.  Access path selection in a relational database management system , 1979, SIGMOD '79.