Analyzing Plan Diagrams of XQuery Optimizers

Abstract

The automated optimization of declarative user queries is a classical hallmark of database technology. XML, with its support for deep data hierarchies and powerful query operators, including regular expressions and sibling axes, renders the query optimization challenge significantly more complex. In this paper, we analyze the behavior of industrial-strength XQuery optimizers using the notion of “plan diagrams”, which had hitherto been applied solely to relational engines. Plan diagrams visually characterize the optimizer’s query plan choices on a parametrized query space, and extending them to the XML environment requires redesigned definitions of the parameters and the space. Through a comprehensive set of experiments on a variety of popular benchmarks, we demonstrate that XQuery plan diagrams can be significantly more dense and complex as compared to their relational counterparts. Further, they are more resistant to “anorexic reduction”, requiring substantially larger cost-increase thresholds to achieve this objective. These results suggest that important research challenges remain to be addressed in the development of effective XQuery optimizers.