Hammerspace has announced a breakthrough IO500 10-Node Production result that establishes a new era for high-performance data infrastructure. For the first time, a fully standards-based architecture — standard Linux, the upstream NFSv4.2 client, and commodity NVMe flash — has delivered a 10-node Production fully reproducible IO500 result traditionally achievable only by proprietary parallel filesystems.
This result is the first IO500 Production benchmark demonstrating indisputable proof that standards-based Linux and NFS can meet the extreme performance requirements of high-performance computing (HPC) and artificial intelligence (AI) workloads — without proprietary client software, specialized networking stacks or complex parallel filesystem infrastructure.
A Milestone Moment for Data Platforms — As Transformative as Linux Was for Compute
In the late 1990s, researchers like Dr. David Bader, Distinguished Professor and founder of the Department of Data Science in the Ying Wu College of Computing and Director of the Institute for Data Science at New Jersey Institute of Technology, transformed the HPC world by proving that Linux-based clusters, built on Linux and on commodity components, could rival proprietary supercomputers. That work transformed HPC architecture then, and machine learning (ML) and AI architectures of the future, ultimately making Linux the standard powering nearly every powerful compute environment on Earth.
This vision laid the foundations for the AI architectures that are emerging even today. Hyperion Research estimates that “over $300 billion in revenue has been generated from selling supercomputers. This represents a sizable economic gain, especially since the use of these systems generated research valued at least ten times over the purchase price. While it is difficult to fully measure the value that supercomputers have generated, even looking at just automotive, aircraft, and pharmaceuticals, supercomputers have contributed to products valued at more than $100 trillion over the last 25 years.”
Hammerspace’s IO500 achievement represents the next chapter of that evolution, this time in the data layer.
Just as Linux revolutionized compute architecture, the combination of standards-based Linux and pNFS is now proving it can revolutionize high-performance data architecture for HPC and AI.
The First Architecture That Meets the Demands of Both HPC and AI
This achievement marks the industry’s proof that open, interoperable infrastructure can deliver the performance required by AI and HPC workloads without proprietary lock-in.
HPC environments have traditionally relied on deep institutional expertise to operate complex proprietary filesystems, but the rapid rise of AI has changed the landscape. AI is scaling far faster — across enterprises, cloud providers, sovereign AI platforms, service providers and thousands of new data-intensive applications — and it is impossible to meet this demand with architectures that require niche expertise to deploy and maintain. Every systems administrator already knows how to operate Linux and NFS; however, very few have the specialized knowledge required for legacy parallel file systems. As AI infrastructure becomes mainstream, organizations need HPC-class performance delivered through tools and protocols familiar to the broader IT community. This IO500 result proves that the performance required for both HPC and AI can now be achieved using standard Linux, standard NFS and widely understood operational models, finally aligning extreme performance with the scale and accessibility the AI industry demands.
Standards-Based Architecture, Industry-Leading Performance
The submission by Samsung, leveraging the Hammerspace Data Platform, achieved the fastest standards-based IO500 10-Node Production result ever recorded. Hammerspace not only contributes a substantial number of the capabilities into Linux for pNFS workloads, but its Data Platform is engineered and designed from the ground up to capitalize on these upstream performance enhancements in the Linux kernel.
Unlike traditional storage platforms and legacy parallel file systems that treat Linux as a compatibility layer or pNFS as an added-on interface, Hammerspace’s architecture is built directly on top of — and actively contributes to — the same NFSv4.2 and pNFS innovations driving modern HPC and AI performance. This deep alignment uniquely allows Hammerspace to take immediate advantage of new capabilities such as lower-latency I/O paths, advanced client-side parallelism and improved failover logic, translating Linux’s ongoing advancements directly into real-world application speedups. As a result, organizations can benefit from cutting-edge performance improvements in standard Linux distributions without deploying proprietary clients or rearchitecting their infrastructure.
Unlike legacy parallel file systems that rely on complex, vendor-specific clients, the Samsung’s Hammerspace submission used:
- Standard RHEL/Ubuntu Linux
- Standard upstream NFSv4.2 (pNFS) client
- Standard NVMe SSDs from Samsung
- Standard IP-over-InfiniBand
- Standard server platforms
- Hammerspace’s standards-based parallel global file system leveraging the pNFS client
In the submission, there was no proprietary client, no custom kernel modules and no exotic parallel file system used.
Modern HPC and AI workloads can now run at elite speeds using standards-based infrastructure and data architectures.
Upstream Linux Innovation Unlocks New Performance
The step-function improvement achieved during the time between the ISC25 and SC25 events is the result of:
- Enhanced pNFS Flexible File layout parallelism
- Upstream NFS client improvements contributed by Hammerspace
- Upstream NFS server improvements that avoid page cache contention, allowing improved sustained performance and reduced resource utilization, contributed by Hammerspace
- File-level objective-based policy optimizations
- Latency reductions and throughput gains in metadata access
- High-performance NVMe data placement managed through the Hammerspace global file system
These enhancements strengthen the entire Linux ecosystem — echoing the transformative Linux HPC contributions of the late 1990s and early 2000s.
Hammerspace’s Top-10 IO500 performance is more than a benchmark victory. It is the first empirical proof that standards-based Linux and NFS can power high-performance data systems at the top of the HPC and AI stack.
Linux democratized supercomputing and now standards-based data infrastructure is positioned to democratize high-performance storage — and reshape the future of global-scale computing.
Hammerspace Breaks IO500 Barriers: First Standards-Based Linux + NFS System To Achieve True HPC-Class Performance
Posted in Commentary with tags Hammerspace on November 21, 2025 by itnerdHammerspace has announced a breakthrough IO500 10-Node Production result that establishes a new era for high-performance data infrastructure. For the first time, a fully standards-based architecture — standard Linux, the upstream NFSv4.2 client, and commodity NVMe flash — has delivered a 10-node Production fully reproducible IO500 result traditionally achievable only by proprietary parallel filesystems.
This result is the first IO500 Production benchmark demonstrating indisputable proof that standards-based Linux and NFS can meet the extreme performance requirements of high-performance computing (HPC) and artificial intelligence (AI) workloads — without proprietary client software, specialized networking stacks or complex parallel filesystem infrastructure.
A Milestone Moment for Data Platforms — As Transformative as Linux Was for Compute
In the late 1990s, researchers like Dr. David Bader, Distinguished Professor and founder of the Department of Data Science in the Ying Wu College of Computing and Director of the Institute for Data Science at New Jersey Institute of Technology, transformed the HPC world by proving that Linux-based clusters, built on Linux and on commodity components, could rival proprietary supercomputers. That work transformed HPC architecture then, and machine learning (ML) and AI architectures of the future, ultimately making Linux the standard powering nearly every powerful compute environment on Earth.
This vision laid the foundations for the AI architectures that are emerging even today. Hyperion Research estimates that “over $300 billion in revenue has been generated from selling supercomputers. This represents a sizable economic gain, especially since the use of these systems generated research valued at least ten times over the purchase price. While it is difficult to fully measure the value that supercomputers have generated, even looking at just automotive, aircraft, and pharmaceuticals, supercomputers have contributed to products valued at more than $100 trillion over the last 25 years.”
Hammerspace’s IO500 achievement represents the next chapter of that evolution, this time in the data layer.
Just as Linux revolutionized compute architecture, the combination of standards-based Linux and pNFS is now proving it can revolutionize high-performance data architecture for HPC and AI.
The First Architecture That Meets the Demands of Both HPC and AI
This achievement marks the industry’s proof that open, interoperable infrastructure can deliver the performance required by AI and HPC workloads without proprietary lock-in.
HPC environments have traditionally relied on deep institutional expertise to operate complex proprietary filesystems, but the rapid rise of AI has changed the landscape. AI is scaling far faster — across enterprises, cloud providers, sovereign AI platforms, service providers and thousands of new data-intensive applications — and it is impossible to meet this demand with architectures that require niche expertise to deploy and maintain. Every systems administrator already knows how to operate Linux and NFS; however, very few have the specialized knowledge required for legacy parallel file systems. As AI infrastructure becomes mainstream, organizations need HPC-class performance delivered through tools and protocols familiar to the broader IT community. This IO500 result proves that the performance required for both HPC and AI can now be achieved using standard Linux, standard NFS and widely understood operational models, finally aligning extreme performance with the scale and accessibility the AI industry demands.
Standards-Based Architecture, Industry-Leading Performance
The submission by Samsung, leveraging the Hammerspace Data Platform, achieved the fastest standards-based IO500 10-Node Production result ever recorded. Hammerspace not only contributes a substantial number of the capabilities into Linux for pNFS workloads, but its Data Platform is engineered and designed from the ground up to capitalize on these upstream performance enhancements in the Linux kernel.
Unlike traditional storage platforms and legacy parallel file systems that treat Linux as a compatibility layer or pNFS as an added-on interface, Hammerspace’s architecture is built directly on top of — and actively contributes to — the same NFSv4.2 and pNFS innovations driving modern HPC and AI performance. This deep alignment uniquely allows Hammerspace to take immediate advantage of new capabilities such as lower-latency I/O paths, advanced client-side parallelism and improved failover logic, translating Linux’s ongoing advancements directly into real-world application speedups. As a result, organizations can benefit from cutting-edge performance improvements in standard Linux distributions without deploying proprietary clients or rearchitecting their infrastructure.
Unlike legacy parallel file systems that rely on complex, vendor-specific clients, the Samsung’s Hammerspace submission used:
In the submission, there was no proprietary client, no custom kernel modules and no exotic parallel file system used.
Modern HPC and AI workloads can now run at elite speeds using standards-based infrastructure and data architectures.
Upstream Linux Innovation Unlocks New Performance
The step-function improvement achieved during the time between the ISC25 and SC25 events is the result of:
These enhancements strengthen the entire Linux ecosystem — echoing the transformative Linux HPC contributions of the late 1990s and early 2000s.
Hammerspace’s Top-10 IO500 performance is more than a benchmark victory. It is the first empirical proof that standards-based Linux and NFS can power high-performance data systems at the top of the HPC and AI stack.
Linux democratized supercomputing and now standards-based data infrastructure is positioned to democratize high-performance storage — and reshape the future of global-scale computing.
Leave a comment »