Multidimensional mapping and indexing of XML

Abstract

We propose a multidimensional approach to store XML data in relational database systems. In contrast to other efforts we suggest a solution to the problem using established database technology. We present a multidimensional mapping scheme for XML and also thoroughly study the impact of established and commercially available multidimensional index structures (compound B-Trees and UB-Trees) on the performance of the mapping scheme. In addition, we compare our multidimensional mapping to other known mapping schemes. While studying the performance we have identified projection and selection to be fundamental parts of a typical query on XML documents. Our measurements show that projection and selection are orthogonal and require special multidimensional index support to be processed efficiently.