National Research Platform - Nautilus
The National Research Platform (NRP) is a comprehensive, heterogeneous, and nationally distributed open system that features a variety of computing resources. Managed jointly by over 50 institutions, led by researchers and cyberinfrastructure professionals primarily at UC San Diego, NRP is designed to facilitate a wide range of data science, simulations, and machine learning or artificial intelligence applications.
System Access
Access to the NRP is primarily through the Nautilus platform, a hypercluster designed for running containerized big data applications using Kubernetes (K8S). To gain access to Nautilus, you can authenticate using CI Login to login using Clemson credentials, but must be added to a namespace to make use of the cluster. For details on the process, please see the Getting Started documentation.
Job Management
NRP employs Kubernetes for job scheduling, monitoring, and management, supporting a range of containerized workloads. Users can manage resources and view logs using the kubectl command-line tool, facilitating detailed control over deployed applications and services.
Nautilus Cluster
The centerpiece of NRP's infrastructure is the Nautilus cluster, a distributed system of approximately 300 nodes, which is anticipated to expand substantially. Nautilus is equipped to host numerous network testing and data transfer nodes used by initiatives like the Open Science Grid (OSG).
Features of Nautilus
-
Computing Power: Nautilus is a potent distributed computing system featuring CPUs, GPUs, and FPGAs in two subsystem configurations, high-performance FP64/FPGA and FP32-optimized systems.
-
Total Computing Cores: The system comprises about 10,000 CPU cores and an expanding array of GPU resources.
-
Data Transfer Nodes: Nautilus includes 100Gb/s data transfer and test nodes, not accounting for additional PNRP Cat-II facilities.
-
Storage: Nautilus supports PRP with 5PB of Ceph Storage, not including another 5PB from PNRP.
For more information, please visit the Nautilus documentation.