ET&S Strategic Communications

In FY24, Enterprise Technology & Services added a new GPU node with NVIDIA H100 cards to the Premise High-Performance Compute Cluster (HPC), a significant enhancement for artificial intelligence and machine learning analysis. The integration with the high-performance storage purchased in FY24 will help boost the HPC's capabilities.

The Premise High-Performance Compute Cluster (HPC) is a robust infrastructure used by over 70 labs for research purposes and is accessible to faculty across all institutions. It underwent significant enhancements with the installation of a new GPU node in May 2024, which included 4 NVIDIA H100 cards optimized for artificial intelligence and machine learning analysis. This addition leverages the high-performance storage acquired in FY24, further augmenting the HPC's capabilities.

The HPC cluster previously faced challenges with its storage solutions, which were either unusable or inadequate. To address this, ET&S introduced a new storage system, Pixstor Storage, in February 2024, along with a support contract to ensure performance and stability. This system accommodates all HPC use cases and sustains long-term usage, benefiting researchers across all institutions.

The HPC's infrastructure is managed by the Data Center Operations team, which supports all on-premises computing and storage, including installing and supporting HPC clusters for research. The team ensures the smooth operation of data centers located at Durham, Keene, Plymouth, Concord, and Manchester, covering various services from HVAC and electrical to networking and security.

The HPC will soon also be accessible via the "Jupyter Hub" web application, which will launch in July 2024. This application will enable users to execute code in data science languages and integrate natural language processing into machine learning, facilitating the creation of local automated workflows.

Overall, the Premise HPC is a critical asset for the research community, providing the necessary computational power and infrastructure to support advanced research and analysis.