Distributed Infrastructure
Over 400 nodes at more than 70 sites on three continents, running one Kubernetes API. Institutions contribute hardware and keep priority on it; everyone else gets opportunistic access to the rest.
What distribution buys you
Standardized environments
Control software and configurations across sites using containerized images for consistency and reproducibility.
Industry-standard orchestration API
A standardized Kubernetes API and resource model make it easy to develop, deploy, and integrate services using familiar primitives (Deployments, Services, ConfigMaps).
Fault tolerance
Distributed sites reduce single points of failure so experiments and services remain available even when individual nodes or even entire sites go offline.
Data locality
Keep data close to compute to reduce latency and improve throughput for large datasets and real-time experiments.
Scalability
Grow your workloads across multiple sites to access more CPU, GPU, and storage resources on demand.
Secure networking & isolation
Encrypted site-to-site links, network segmentation, and tenant isolation protect traffic and reduce the risk of lateral movement.
How the Open Science Data Federation builds on the NRP
OSDF hosts its national data-distribution network on NRP sites, without running the systems underneath it.
Distributed caches
OSDF caches sit at multiple NRP sites, so data is served from near the researcher rather than from one origin.
Local origins
Origins run on hardware near the data source, owned by the data provider, while the NRP handles service hosting.