Using Characteristics of Computational Science Metadata for Efficient XML Storage in a Relational Database

Computational science communities are generating an ever-increasing volume of data products which are tracked and cataloged using metadata communicated via one or more community XML schemas. We propose a hybrid approach to storing metadata in a relational database that draws heuristics from computational science discovery, and show that the approach outperforms the well known inlining approach. Additionally, because the hybrid approach is loosely coupled to the XML schema, it is more adaptable to schemas across a broader spectrum of scientific domains.