Storing XML (with XSD) in SQL databases: interplay of logical and physical designs

In this paper, we examine the interplay of logical and physical design, and experimentally demonstrate that: (1) solving the logical mapping and the physical design problem independently leads to a suboptimal solution; (2) taking into account the physical design space impacts the space of logical mapping. Specifically, well-known outlining and inlining mapping options are rendered unnecessary because they are functionally subsumed by two physical design options: indexes and vertical partitioning. We propose a search algorithm that judiciously explores the extreme large combined space of logical and physical design. The algorithm only searches the XSD-specific logical design options and uses heuristics to further prune the search space. We experimentally compare the quality (in terms of the time to execute the query workload on resulting design) and efficiency (in terms of the search time) of our algorithm with known algorithms as well as a default XSD based mapping and an Edge-Table Mapping that does not use XSD on both real and synthetic data.