CC* Compute: Koa - A High Performance and Flexible Research Computing Resource

点击翻译按钮获取中文摘要

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).The University of Hawaii (UH) aims to establish the CC* Compute: Koa - A High Performance and Flexible Research Computing Resource to support computationally intensive research, education and practice across the UH ten-campus system. The project aims to install a compute cluster Koa, which is architected in response to current resource constraints, particularly scratch storage. The cluster hardware includes scratch storage of 750TB, 2 Graphics Processing Unit (GPU) nodes comprising 10x Nvidia A4000 cards, and a total of 384 CPU cores across 8 compute nodes. Koa provides UH faculty, researchers and students state-of-the-art computational resources focused towards machine learning, artificial intelligence and large scale simulation. The Koa computing cluster is a shared inter-campus resource available to all UH researchers. Koa focuses to support researchers in the specialties of astronomy, atmospheric science, ocean science, microbiome science, and computer & data science, across UH's ten-campus system. Koa resources enable researchers to scale up research to process larger datasets and models in addition to accelerating existing workflows through the new advanced architecture that ties extremely fast storage to the compute resources over a high-speed network, which enables a shorter time to result. The strategic partnership with the Open Science Grid allows for efficient usage of unused Koa resources by the national research community as well as a gateway to a national compute platform for Koa users. Additionally, the Koa resource aids hands-on training in data science and computational science for the next generation of researchers and data scientists through partnership with the Hawaii data science institute and local community for workshops and classroom access.Koa enables a larger scope and scale of analysis by providing 750TB of high-speed Lustre parallel scratch storage capacity accessible from all compute and GPU accelerated resources. Koa’s 200Gb HDR infiniband network enables Koa’s Lustre to achieve I/O speeds up to 96Gb/s and provide increased computational throughput for I/O heavy workflows. Koa’s external network connections increase data transfer speeds to national, commercial cloud and academic resources via the combination of 100Gb/s data transfer node connection and high-speed parallel file system, enhancing end-to-end big data workflows. Koa also provides virtualized infrastructure to support specialized computational use cases such as: immersive analytics and science gateways. The strategic partnership with the Open Science Grid allows for harvesting unused cycles on Koa by the national research community as well as a gateway to national compute platform for Koa users. The integration of Koa’s high speed file system with the regional Jestream2 NSF Cloud infrastructure hosted at UH allows researchers to easily span computing environments and modalities between the cloud and local Koa computing resources to support new deep learning and artificial intelligence workflows, visualizations and applications.This project is funded through the collaborative efforts of the Office of Advanced Cyberinfrastructure (OAC) and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.