Vertical partitioning in object oriented databases

Abstract

Database partitioning is a useful technique for performance improvement in parallel and dis-tributed environments. Reported ways of partitioning object oriented database (ODB) classes include vertical, horizontal, mixed, path and method induced partitioning. This thesis provides method transformations and cost analysis of vertical partitioning in parallel object oriented database systems (POBSs). A taxonomy for vertical partitioning schemes is proposed and it is used qualitatively to justify the need for new partitioning algorithms. A partitioning algorithm and support for transparency to applications in the partitioned database are two essential components of a vertical-partitioning processor. The thesis presents a unified view of the vertical partitioning problem which incorporates these two components.