Video: Nautilus and the National Research Platform

About this video

Nautilus is a 250-node, 20-campus, highly distributed, centrally-managed computer cluster mainly serving researchers’ advanced GPU needs. Launched by five NSF CISE awards to UC San Diego, Nautilus features 1,000 GPUs, 10,000 CPU cores, and 9 PB of nationally distributed fast storage, including 100s of researcher-added GPU nodes. Researchers on 30 campuses run software in containers, orchestrated by Kubernetes, bursting from their laptop environments to dozens or 100s of GPUs at a time. Nautilus exploits the Science DMZs previously built with NSF funding on over 100 US campuses. Crucially, campuses contribute power, rack space, and 10-100G networking for Nautilus nodes.