Data curation with a focus on reuse

Abstract

A dataset from the field of High Performance Computing (HPC) was curated with the focus on facilitating its reuse and to appeal to a broader audience beyond HPC specialists. At an early stage in the research project, the curators gathered requirements from prospective users of the dataset, focusing on how and for which research projects they would reuse the data. Users needs informed which curation tasks to conduct, which included: adding more information elements to the dataset to expand its content scope; removing personal information; and, packaging the data in a size, a format, and at a frequency of delivery that are convenient for access and analysis purposes. The curation tasks are embedded in the software that produces the data, and are implemented as an automated workflow that spans various HPC resources, in which the dataset is generated, processed and stored and the Texas ScholarWorks institutional repository, through which the data is published. Within this distributed architecture, the integrated data creation and curation workflow complies with long-term preservation requirements, and is the first one implemented as a collaboration between the supercomputing center where the data is created on ongoing basis, and the University Libraries at UT Austin where it is published. The targeted curation strategy included the design of proof of concept data analyses to evaluate if the curated data met the reuse scenarios proposed by users. The results suggest that the dataset is understandable, and that researchers can use it to answer some of the research questions they posed. Results also pointed to specific elements of the curation strategy that had to be improved and disclosed the difficulties involved in breaking data to new users.