Techniques for integrating tertiary storage with object-relational database management systems

The increasing popularity of incorporating new and complicated data types into database systems has prompted the rapid development of object-relational database technology over the past few years. One of the most important features of an object-relational database system is the ability of such systems to store and manipulate large, less-structured multimedia data types including images, text, video and audio. Instances of these data types are typically hundreds of thousands times larger than the traditional fields in a relational table. Consequently, databases that contain instances of these data types tend to be much larger than normal. This increase in database size makes the use of low-cost, high-capacity tertiary storage devices as an alternative storage medium attractive. However, most existing DBMS can only process data stored on conventional secondary storage devices and treat tapes primarily as backup or archive devices since their random access times are very long compared to disks.; This thesis presents a set of techniques designed for the seamless integration of tape-based tertiary storage devices into an object-relational database system for storing and manipulating large objects. These techniques are centered on the simple, yet powerful, idea of query pre-execution. Query pre-execution is designed to gather accurate run-time access patterns for tertiary storage accesses so that these accesses can be optimized in order to reduce the number of random accesses performed during query execution. Practical issues such as large-object representation, tertiary block size selection, tertiary storage access scheduling, batch query execution, and tertiary storage volume management are investigated as part of this thesis. The thesis also describes the design and implementation of these techniques in Paradise, a parallel object-relational database system designed to efficiently process a variety of new data types including spatial data, satellite imagery, video, and text. A performance evaluation of these proposed techniques is included in the thesis to demonstrate their effectiveness.