Capturing Service Versioning in Provenance Trace to Support Reproducibility

Abstract

Reproducibility has long been a cornerstone of science. Underpinning reproducibility is provenance, which has the potential to provide scientists with a complete understanding of data generated in e-experiments, including the services that were produced and consumed. A key to reproducibility is the provenance model: a data model that structures information about an e-experiment. When all the entities in the experiment have been identified, they must be captured and recorded as a provenance trace. The provenance trace gives information about the actual execution of an experiment. Therefore, in running an experiment, the creation of the final results that are derived from the input data are documented in a provenance trace. This paper describes in greater detail the conceptualization of an experiment using the Open Provenance Model (OPM). As Open Provenance Model (OPM) is the provenance model standard, this paper explores whether the OPM is able to describe an experiment sufficiently precisely so as to support reproducibility. The paper also addresses the issue of how to ensure that the versions of services involved in the experiment can remain available, as service versioning is part of essential requirements in reproducibility.