Experimental Research Reproducibility and Experiment Workflow Management

Research reproducibility is an important factor to support the full research life cycle; this is especially important for experimental research, where it is required to also reproduce the whole experiment environment and equipment setup. This paper presents the methodology and solution proposed in the SLICES Research Infrastructure to enable research reproducibility in modern digital technologies for complex and large scale experimentation. The paper provides a short overview of existing research and approaches for experimental research reproducibility, generally including git based experiments deployment and operation, Jupyter Notebook and Common Workflow Language (CWL) for workflow management. The paper describes the framework and approaches taken in the SLICES-RI that also address research environment provisioning on demand with the Platform Research Infrastructure as a Service (PRIaaS) and data management infrastructure to ensure data quality and support effective data sharing.

I. Introduction

Modern research is increasingly multidisciplinary and data driven, emphasizing the need for effective communication between researchers and data sharing. The Open Science initiative and movement among research communities can address these needs and increase the overall efficiency of scientific and technology research [1]. Open Science is strongly supported by policy development and funding bodies in Europe and a mandatory requirement in the current Horizon Europe programme. In recent years, major initiatives and projects to create the foundation and e-Infrastructure for Open Science have been funded in Europe (in the framework of the past Horizon 2020 and current Horizon Europe programmes). This includes the currently operational OpenAIRE [2] and Zenodo [3] services, the development of the European Open Science Cloud (EOSC) [4], all support the FAIR (Findable, Accessible, Interoperable, Reusable) data principles [5]. The FAIR data principles were initially proposed for research data management primarily targeted at consistent metadata management. However, recent initiatives by the Research Data Alliance (RDA) [6] are extending FAIR principles to scientific software as a digital or data object.