On the Difficulty of Finding Optimal Relational Decompositions for XML Workloads: A Complexity Theoretic Perspective

Abstract

A key problem that arises in the context of storing XML documents in relational databases is that of finding an optimal relational decomposition for a given set of XML documents and a given set of XML queries over those documents. While there have been a number of ad hoc solutions proposed for this problem, to our knowledge this paper represents a first step toward formalizing the problem and studying its complexity. It turns out that to even define what one means by an optimal decomposition, one first needs to specify an algorithm to translate XML queries to relational queries, and a cost model to evaluate the quality of the resulting relational queries. By examining an interesting problem embedded in choosing a relational decomposition, we show that choices of different translation algorithms and cost models result in very different complexities for the resulting optimization problems. Our results suggest that, contrary to the trend in previous work, the eventual development of practical algorithms for finding relational decompositions for XML workloads will require judicious choices of cost models and translation algorithms, rather than an exclusive focus on the decomposition problem in isolation.

Research supported in part by NSF grants CSA-9623632, ITR-0086002, CCR- 9634665 and CCR-0208013