SLICES Data Management Infrastructure for Reproducible Experimental Research on Digital Technologies

This paper presents the ongoing research effort related to the design of the Data Management Infrastructure (DMI) to support experimental research on digital technologies with application to the ESFRI SLICES scientific instrument. We consider the experiment documentation and data collection across the whole continuum of access network, IoT, edge, cloud, and data processing workflow. The paper includes the requirements analysis for DMI to enable research reproducibility of complex and large-scale experimentation. We provide an analysis of data collected and processed in SLICES and explain approaches and solutions used in SLICES for experimental research reproducibility, primarily based on the plain orchestration service and supported by metadata collection tools. The proposed multi-layer DMI includes: data (storage) access, data processing, data ingest, experiment management, and virtual research environment. The paper also provides recommendations for the selection of existing standards and tools for data and metadata management, in particular those developed by EOSC and supported by the RDA community to ensure wide compatibility and integration.

I. Introduction

Wider adoption of Open Science requires a modern research infrastructure and scientists to pay more attention to consistent data management in order to support effective data sharing and communication between researchers [1]. Introducing FAIR data principles and ongoing development and implementation of supporting standards, frameworks, and tools in recent years, significantly improved the possibility for sharing research data and research results, targeting research reproducibility, sharing data, or other publishable research results via the popular Open Access or self-archiving services OpenAIRE [2] and Zenodo [3]. The European Open Science Cloud (EOSC) [4] provides the federated data sharing infrastructure. Recent developments such as RO Crate [5,6] have the potential of supporting complex research objects and their evolution. This is especially important for experimental research reproducibility that requires documenting a large volume of information related to the experiment setup, workflow, input data, measurement data [7].