· Events · 5 min read
NRP at SC25
National Research Platform showcases distributed cyberinfrastructure and networking innovations at Supercomputing 2025 in St. Louis
The National Research Platform (NRP) had a strong presence at this year’s ACM/IEEE Supercomputing Conference (SC25), held in St. Louis, Missouri. Throughout the week, NRP was featured across tutorials, technical sessions, live demos, and booth engagements — each highlighting how AI, networking, and modern orchestration technologies can come together to push the boundaries of scientific computing.
📚 Tutorial: AI and Scientific Research Computing with Kubernetes
NRP colleagues led a tutorial at SC25 on AI and Scientific Research Computing with Kubernetes. Kubernetes has emerged as the leading container orchestration solution (maintained by the Cloud Native Computing Foundation) that works on resources ranging from on-prem clusters to commercial clouds. Kubernetes capabilities are available on Expanse, Voyager, and Prototype National Research Platform (PNRP) Nautilus clusters at SDSC. These clusters support AI and scientific computing research workloads.
Recently there has also been rapid growth in the use of AI resources for educational purposes. Several institutions have incorporated LLMs into their curriculum, leveraging Nautilus services and resources. This tutorial aimed to educate AI and computational science researchers on the capabilities of Kubernetes as a resource management system, compared with traditional batch systems; provide information on useful IO/storage options, and optimal use strategies for AI workloads; and demonstrate the use of Kubernetes-based solutions integrating LLM inference use for classroom use via JupyterHub.
Attendees received an overview of the Kubernetes architecture, typical job and workflow submission procedures, use various storage options, run AI and scientific research software using Kubernetes using both CPU and GPU resources, learn about optimal I/O strategies for AI, and run examples leveraging LLM inference services on Nautilus. Theoretical information was paired with hands-on sessions operating on the PNRP production cluster Nautilus.
🌐 NRP in SCinet
NRP was directly integrated into the conference’s own infrastructure through SCinet, with SCinet’s HPC cluster temporarily federated into the National Research Platform. This live integration enabled NRP services, Kubernetes orchestration, and AI-driven workflows to operate on conference-hosted resources in real time — a powerful example of on-demand, federated cyberinfrastructure in action.
NRP’s participation in SCinet, the conference’s dedicated high-speed research network, demonstrated the platform’s capabilities in wide-area networking and distributed infrastructure. Through SCinet, NRP showcased:
- Federated infrastructure connecting multiple sites across the United States
- Programmable networking capabilities using P4, SmartNICs, and FPGA technologies
- Real-time data processing and in-network computing demonstrations
- Cross-site collaboration tools and workflows for distributed scientific computing
- Live SCinet-connected Nautilus nodes enabling seamless execution of high-performance, interactive workloads
📚 Network Research Exhibitions
NRP presented several Network Research Exhibitions (NREs) at SC25, demonstrating cutting-edge research in distributed cyberinfrastructure, programmable networking, and in-network computing:
NRE 119: The National Research Platform and SCinet
Enabling Live, Multi-Institutional Scientific AI/ML and HPC Workflows
The San Diego Supercomputer Center and the Prototype National Research Platform (NSF Award 2112167 Category II) provide the National Research Platform, a community-owned platform connecting researchers and educators to accelerate innovation and share resources. Supported by over 70 institutions, the NRP provides access to cutting-edge technologies in AI, high-performance computing, data storage, and networking using the Nautilus Kubernetes Cluster, which directly incorporated SCinet-connected compute nodes on the exhibit floor at SC25 to enable seamless execution of high-performance, interactive workloads.
NRE 120: Live High-Precision Per-Packet Kubernetes Data Telemetry
From NRP to StarLight via ESnet SENSE
This demonstration linked the National Research Platform (NRP) Kubernetes cluster resources at the San Diego Supercomputer Center (SDSC) to the StarLight facility in Chicago via the ESnet SENSE Orchestrator. Using dedicated 100Gbps direct links integrated into the AutoGOLE global topology, NRP transmitted live Kubernetes network data packets to P4-programmable Tofino2 SmartNIC endpoints deployed at StarLight for advanced packet inspection, telemetry, and analysis. Each packet originating from SDSC, where numerous NRP resources and Kubernetes workloads operate, is routed over SENSE using orchestrated L2 paths with a dedicated QoS VLAN tag. Upon arrival at StarLight, programmable P4 logic running on Tofino2 returns header-only data stream to a Xilinx Alveo FPGA running P4 and DPDK acceleration for high-throughput packet processing and offload, ingesting it into a Kafka streaming database in real time, enabling live monitoring and analysis of Kubernetes cluster network behavior and workload communication patterns.
NRE 121: Real-Time In-Network Machine Learning and P4 Testbed Deployment
On FPGA SmartNICs, DPUs, and Switches
The San Diego Supercomputer Center and National Research Platform are deploying a large-scale programmable testbed integrating P4-programmable SmartNICs, DPUs, and switches for real-time in-network computing research. This infrastructure includes seven 1U DC-powered measurement and monitoring servers, known as Interactive Global Research Observatory Knowledgebase (IGROK) nodes, supplied through the PacWave NSF Award 2029306. Each IGROK node is equipped with BlueField-2 DPUs featuring 2×100 Ethernet interfaces and is integrated into the AutoGOLE/SENSE topology, enabling distributed monitoring and control.
NRE 122: Agentic AI with Qualcomm Cloud AI 100 Ultra Cards
For HPC Cluster Management and Resource Provisioning
The San Diego Supercomputer Center and the Prototype National Research Platform (NSF Award 2112167 Category II) provide the National Research Platform, a community-owned platform connecting researchers and educators to accelerate innovation and share resources. Supported by over 70 institutions, the NRP provides access to cutting-edge technologies in AI, high-performance computing, data storage, and networking using the Nautilus Kubernetes Cluster, which directly incorporated SCinet-connected compute nodes on the exhibit floor at SC25 to enable seamless execution of high-performance, interactive workloads.
This Network Research Exhibition was honored with the SC25 Spirit of Innovation Award.
This blog post was edited using NRP’s publicly hosted LLMs. Learn more.
