Object-relational databases extend the capabilities of the relational databases by allowing definition of new data types and methods to operate on these data types while retaining most of the relational model semantics. These extensions together with ever increasing size of the data that the database systems are expected to handle pose new challenges in query processing. Parallel object-relational database systems address these issues by storing and manipulating the data across multiple processors. In this paper, we propose a new technique to speed up method execution on large objects. It modifies I/O access pattern of the query to achieve better I/O performance. This is achieved by selectively migrating and reordering the method execution tasks. This technique part of a larger I/O optimization scheme based on collective I/O and provides as much as 600\% improvement in query execution performance.

Gzipped Postscript version of the paper