HPC & GPU Resources for AI/ML Augmented Biomedical Research

Project Summary/Abstract: Fred Hutchinson Cancer Center respectfully requests funds to replace its aging high-performance computing (HPC) cluster, aligning with technological advances and research computing trends from 2020 to 2025. The existing 212-node, 7,128-core HPC cluster, established in 2020 is reaching the end of its supported life and now requires replacement to facilitate biomedical research by integrating GPUs, leveraging their capabilities in artificial intelligence and machine learning. The proposed HPC system, slated for deployment in an updated space within Fred Hutch's datacenters, aims to replace this existing HPC system by adding 130 new nodes (totaling 7,799 cores). This system will culminate in a state-of-the-art 130-node, 7,799-core, 192 TB RAM HPC system with 1,753,088 CUDA cores, signifying a 9% surge in core count, an 97% boost in GPU processing power, and a 40% increase of memory capacity compared to the existing infrastructure. These enhanced computational capabilities are poised to catalyze deep and efficient analysis across a spectrum of biomedical research domains, facilitated by an increased emphasis on GPU-accelerated methodologies. Biomedical investigations, ranging from cancer eradication to infectious disease modeling, will benefit significantly from this computational enhancement. The core user group comprises no less than 38 NIH-funded research cohorts engaged in endeavors such as vaccine development, large-scale integrative analytics, and genomic studies, and are now increasingly reliant on GPU-accelerated algorithms for expedited insights. Furthermore, Fred Hutch acknowledges the imperative to meet these escalating computational demands of its user community and plans complimentary upgrades to the environment. Several major users are currently grappling with project delays, while others encounter hardware limitations with the incumbent infrastructure. This proposed upgrade not only addresses immediate needs but also anticipates future requirements, positioning Fred Hutch at the forefront of computational biomedical research. By embracing GPUs and contemporary computing paradigms, Fred Hutch endeavors to empower its researchers with the computational resources necessary for cutting-edge investigations spanning cancer biology, infectious diseases, and beyond, ultimately advancing scientific understanding and therapeutic innovation in the pursuit of improved human health.