CARC prepares for deployment of new high-performance computing cluster

By Mariah Rosales

The Center for Advanced Research Computing (CARC) is gearing up to deploy its latest high-performance computing (HPC) cluster, tentatively set for a summer launch. This state-of-the-art system will provide researchers with cutting-edge computing power, supporting a wide range of computational and data-intensive projects.

The new cluster, named Easley, honors the legacy of Annie Easley, a pioneering computer programmer and mathematician whose contributions to early rocket propulsion and coding have left a lasting impact on the field of computational science.

Hussein Al-Azzawi, CARC IT Manager, says “We are very excited to launch Easley, our new cluster, and bring new cutting-edge technology to support UNM faculty, staff, and students. We hope always to provide resources that allow research to push the boundaries and act as a crucial resource for those looking to utilize the latest technological innovation when conducting their research at UNM.”

The new Dell HPC cluster will replace Wheeler and will give us an estimated performance boost of 7× to 10×, depending on workload type, and a brand-new all-flash IBM parallel filesystem.

“This is a massive performance upgrade in efficiency and raw power, which comes with a fivefold reduction in the number of racks and physical CPUs,” says Al-Azzawi. “This lowers power consumption per compute unit and reduces cooling and maintenance costs while providing modern architecture and supporting newer software optimizations, preparing us to advance in the AI and machine learning age.”

Easley will be a mixed-use Dell hardware CPU/GPU cluster, incorporating H100 GPUs, which are among the most advanced graphics processing units available today. These GPUs are designed to accelerate AI, machine learning, and scientific computing workloads, offering significant performance enhancements over previous generations.

At announcement, the cluster comprised 60 nodes, 3,968 cores, 22.5 TB RAM, and 32 NVIDIA GPUs. Easley’s core network is based on NVIDIA’s 800 Gbps InfiniBand, with 200 Gbps connectivity to the nodes and a 768 TB raw all-flash parallel filesystem.

Accompanying the deployment is an IBM storage solution, ensuring researchers have access to a robust and scalable data storage infrastructure. Additionally, the system utilizes a high-speed InfiniBand connection, allowing for seamless data transfer and communication between computing nodes, optimizing performance for complex simulations and large-scale computational tasks.