
Can standards-based parallel NFS outperform traditional high-performance file systems like Lustre? Recorded live at SC26 in Chicago, Data Unchained host Molly Presley welcomes back Jon Flynn, Director of Applied Systems at Hammerspace, to explore breakthrough IO500 benchmark results and the innovations pushing high-performance computing storage to new speeds. Jon reveals how Hammerspace achieved 135 GB/s throughput from a single server and how testing on Los Alamos National Laboratory's Hellscape supercomputer delivered approximately 304 GiB/s average read performance. The conversation explores client-side striping, Flex Files, metadata optimization and Linux kernel advancements that are transforming how organizations approach HPC and AI infrastructure. Jon also explains why open standards matter, how upstream Linux contributions benefit the broader industry and what it takes to bring experimental performance breakthroughs into production environments.
As AI workloads place increasing demands on storage infrastructure, could parallel NFS provide the performance, flexibility and efficiency organizations need without relying on proprietary file system clients? Discover what these advances mean for the future of supercomputing, enterprise storage and AI-driven research.
Learn more about Hammerspace: https://hammerspace.com
Data Unchained is powered by Hammerspace.
Hosted on Acast. See acast.com/privacy for more information.
Więcej odcinków z kanału "Data Unchained"



Nie przegap odcinka z kanału “Data Unchained”! Subskrybuj bezpłatnie w aplikacji GetPodcast.








