Optimizing XML twig queries in relational systems

Abstract

In this paper, we propose a new approach for optimizing structural-join and twig queries for XML documents more effectively by utilizing the existing facilities of the relational database systems. Our approach is based on an enhanced structure of a compact Dewey-based labeling scheme. Further, we introduce techniques to make RDBMS more tree-aware without the need to modify the system kernel and without sacrificing space by exploiting the XML schema summary. Our techniques are portable and can be applied to any Dewey-based labeling technique. An extensive experimental evaluation confirms the performance benefits of our approach. In particular, we report on extended twig query performance tests between our system and other two well known XML management systems (eXist: native, MonetDB/XQuery: relational-based system).