Internet2 and Cisco Expand Shared AI Computing Capacity with New Deployments

Internet2 and Cisco have expanded their collaboration on the National Research Platform (NRP) by deploying two new AI PODs at the University of Nebraska, Lincoln and Sacramento State. This move aims to increase shared computing capacity for researchers across the platform, particularly in areas where local resources are insufficient.

The AI PODs provide pre-validated compute, networking, and management capabilities that can be deployed as a unit, reducing engineering and integration work required to add new resources to the platform. This standardized approach will enable Internet2 and other NRP contributors to easily expand their capacity in response to growing demand for AI computing.

Each AI POD includes 8 NVIDIA H200 GPUs with 1.1 TB of combined GPU memory, 128 CPU cores, 2 TB of system memory, and 90 TB of NVMe flash for staging data. The deployment is part of Internet2's ongoing efforts to invest in shared research infrastructure and support the growth of AI-enabled discovery across disciplines.

Internet2 operates dedicated Open Science Data Federation nodes across the country, which function as caches to keep datasets closer to compute resources and researchers using them. This distributed platform requires more than just GPUs, with data needing to reach the compute and paths between nodes being measurable.