Hammerspace sets new IO500 record for HPC-class performance using standard Linux, NFS, and NVMe storage

Hammerspace has announced a new IO500 10-node Production benchmark record, achieved in partnership with Samsung, by running its data platform on standard Linux distributions, the upstream Network File System version 4.2 (NFSv4.2) client, and off-the-shelf NVMe solid-state drives. This marks the fastest standards-based 10-node production result in IO500 history and, according to Hammerspace, is the first time a reproducible, production-level score rivaling proprietary parallel file systems has been achieved with a fully standards-based stack.

The IO500 benchmark is widely recognized in high-performance computing (HPC) and artificial intelligence (AI) for evaluating storage system throughput and latency at scale. Hammerspace claims its result demonstrates that standard Linux and NFS can now meet the data performance requirements of demanding AI and HPC workloads—eliminating the need for proprietary clients, specialized networking stacks, or complex parallel file system infrastructures.

For this submission, Hammerspace and Samsung utilized standard Red Hat Enterprise Linux (RHEL) or Ubuntu distributions, upstream NFSv4.2 with parallel NFS (pNFS) support, Samsung NVMe SSDs, standard IP-over-InfiniBand networking, and off-the-shelf server hardware. No proprietary clients, custom kernel modules, or exotic parallel file systems were used. Hammerspace notes that its parallel global file system leverages and actively contributes to Linux kernel improvements in NFS and pNFS, translating upstream innovations—such as enhanced client-side parallelism, improved failover logic, and latency optimizations—directly into end-user performance gains.

Key technical advancements cited in the IO500 result include improvements in pNFS Flexible File layout parallelism, upstream NFS client and server enhancements contributed by Hammerspace, file-level policy optimizations, reduced latency and improved throughput in metadata access, and high-performance NVMe management within the Hammerspace global file system. Hammerspace emphasizes that these enhancements are now part of the standard Linux ecosystem, allowing users to benefit from performance gains without vendor lock-in or the need for specialized operational expertise.

“This IO500 result rewrites long-standing assumptions about what standards-based Linux and NFS are capable of,” said Trond Myklebust, CTO of Hammerspace and Linux NFS client kernel maintainer. “Achieving a leading 10-Node Production score with the Hammerspace parallel global file system using upstream Linux, pNFS and NVMe hardware demonstrates that HPC-class performance no longer requires proprietary clients or specialized file systems. This achievement is a significant moment for the Linux performance community.”

Source: Hammerspace

Get Data Center Engineering News In Your Inbox:

Popular Posts:

695fcac850f073b041e711a2_karman-p-3200 copy
Karman launches 10 MW Heat Processing Unit for giga-scale AI data center cooling
Screenshot
Five AI data centers to reach 1 GW power capacity in 2026, new analysis shows
68e79d30a17eea847251fae6_img-home-product-liquidjet-main
Frore Systems updates LiquidJet direct-to-chip coldplate for 1,950 W NVIDIA Rubin data center GPUs
1765906506220
Tritium launches 800 VDC bidirectional inverter for data centers and renewable energy sites
Screenshot
HC Capital Partners and Herrmann Family Companies plan 1,500-plus-acre Energy Ranch power-linked data center campus in South Texas

Share Your Data Center Engineering News

Do you have a new product announcement, webinar, whitepaper, or article topic? 

Get Data Center Engineering News In Your Inbox: