A federated policy-governed architecture for sovereign European data spaces
TL;DR
This article proposes and analyses a conceptual, policy-governed federated architecture derived from the architectural work of the Horizon Europe NOUS project and maps architectural functions onto major European reference initiatives.
Abstract
European data spaces aim to enable trusted, sovereign and interoperable data sharing across organisational boundaries. Existing reference frameworks address parts of the problem in isolation: GAIA-X concentrates on federated trust and self-descriptions, the International Data Spaces Reference Architecture Model (IDS-RAM) emphasises usage control and the connector pattern, the Data Spaces Support Centre (DSSC) Blueprint consolidates governance vocabulary and Simpl-Open provides a vendor-neutral middleware substrate. The integration of semantic interoperability, policy-driven sovereignty, federated computation across the cloud–edge–high-performance computing (HPC) continuum and decentralised governance within a single deployable architecture remains an open design problem. This article proposes and analyses a conceptual, policy-governed federated architecture derived from the architectural work of the Horizon Europe NOUS project (Grant Agreement No. 101135927). The architecture is organised in five functional layers — data, services, compute, federation and governance — and revolves around five interacting roles: the Data Space Participant (DSP), the Data Space Infrastructure Participant (DSIP), the Data Space Common Services (DSCS) and the companion NOUS Data and Infrastructure Agents (NDA, NIA). A clear separation between Control Plane and Data Plane mediates policy negotiation, contract enforcement, federated learning and infrastructure provisioning across cloud, edge, HPC and quantum back-ends. We illustrate the model through two representative scenarios derived from NOUS use cases — crisis management with heterogeneous geospatial and field data, and cross-organisational federated analytics over heterogeneous compute resources — and we map architectural functions onto major European reference initiatives. The contribution is conceptual: the article does not claim empirical validation, performance benchmarks or formal certification of standards compliance. Limitations and a research agenda — prototyping, federated interoperability testing and formal verification of policy enforcement — are stated explicitly.
