Hammerspace v5.2 boosts AI & HPC with cloud & security gains
Hammerspace has released version 5.2 of its Data Platform software, bringing performance improvements, expanded cloud integration, and enhanced security features. The upgrades are designed to support AI and high-performance computing (HPC) workloads across on-premises, hybrid, and cloud environments.
Performance gains
The updated software delivers a 33.7% increase in the IO500 benchmark score compared to its previous version, published five months earlier. Total bandwidth has doubled, and some sub-tests, such as IOR-Hard-Read, have reported performance improvements exceeding 800%.
Hammerspace attributes a portion of these results to its client-side NFS performance enhancements, which have been contributed to the Linux kernel. These upgrades are intended to accelerate AI and HPC workloads, while remaining standards-based to avoid vendor lock-in or the need for proprietary software on application servers.
The software's compatibility with any storage platform is intended to allow organisations to optimise performance and latency for new workloads, including large-scale AI model training and inference, without migrating data to new storage systems.
Data locality
The latest release introduces Tier 0 affinitization, which automatically aligns data with the optimal servers within a GPU cluster. This feature reduces east-west network traffic and simplifies deployment by eliminating manual configuration. It is enabled by default, aiming to enhance throughput for cluster-based computational tasks.
Share Referrals is another addition to support extreme scalability. This mechanism distributes the namespace across multiple metadata servers to accommodate very large file counts, ensuring linear scalability as data volumes grow in AI and HPC environments.
Security controls
New security capabilities include Kerberos authentication and Labelled NFS support. The latter enables Linux's SELinux and other Mandatory Access Control systems to apply and enforce security labels across NFS, providing more precise data access control. These measures are targeted at sensitive sectors, including research, government, and regulated industries.
Cloud expansion
Hammerspace v5.2 expands its integration with Oracle Cloud Infrastructure (OCI), offering new operational shapes, including bare-metal. Support for OCI Dedicated Regions will follow, aimed at organisations with strict data sovereignty requirements. The integration extends across hybrid and multi-cloud environments, using Hammerspace's S3-connector technology to bridge on-premises environments with GPU-accelerated cloud clusters on OCI, AWS, and Azure.
According to Hammerspace, the system's global namespace allows users and applications consistent access, regardless of how data is distributed or moved between storage types and locations. This architecture is used by organisations such as Meta to move workloads between on-premises data centres and cloud clusters, with data orchestration handled automatically in the background.
"AI is fundamentally changing how organizations interact with their data. Workloads that were once separate are now deeply interconnected, and the data platform must keep pace. The v5.2 advancements strengthen our ability to unify and accelerate data for AI, HPC and enterprise environments without requiring customers to rebuild storage silos or redesign their infrastructure. It marks another important step toward enabling truly AI-ready data everywhere," said Molly Presley, SVP Global Marketing, Hammerspace.
