Shared HPC storage solutions for research at Stanford
TimeTuesday, July 2411am - 11:30am
DescriptionThe purpose of the talk is to show how Lustre is enabling the Stanford Research Computing Center (SRCC) to provide quality HPC storage at scale, with a low TCO. We’ll first present our Lustre-based “cheap and deep” Oak storage system after more than a year of production, where the characteristics of affordability, scalability, and diversity of access protocols/methods are essential. Then we’ll describe our new scratch storage system for Stanford’s Sherlock HPC cluster and how we plan to design a new system architecture by leveraging innovative Lustre features like data-on-MDT.