Using Characteristics of Computational Science Schemas for Workflow Metadata Management

Computational science workflows are generating an ever-increasing volume of data products. Metadata for these workflows is communicated using one or more discipline-specific schemas and is not static but instead is subject to frequent updates and additions. In contrast to general XML data, the unique uses for scientific metadata allow further optimization.We propose a general metadata catalog for storing scientific metadata that is optimized for community science use and communicates metadata as XML using the schemas of scientific domains. In this paper we show that our hybrid approach outperforms the well-known inlining approach to storing XML when applied to scientific metadata.