SNIA Developer Conference September 15-17, 2025 | Santa Clara, CA
Floyd Christofferson
VP Product Marketing
Hammerspace
Christofferson brings more than 25 years of content creation, content management and storage expertise focused on the methods and technologies needed to use and manage massive volumes of data. Most recently, he was CEO of StrongBox Data Solutions, an autonomous data management and data archiving provider. Previously, Christofferson served as CMO at Arcitecta IP Pty Ltd, a comprehensive data management platform company, and as Director, Storage Product Marketing at SGI for its InfiniteStorage products.
Related Sessions
Activating Untapped Tier 0 Storage Within Your GPU- and CPU-based Compute Clusters
The growing complexity and extended context lengths for inferencing workloads in AI projects have added a costly level of complexity to implementing such initiatives, resulting in increasing I/O required to push data back and forth across the network. This leads organizations to need higher performing storage & faster networks to feed the compute clusters, and get better utilization of their infrastructure.
However, hidden in plain sight within existing GPU- and CPU-based compute platforms lies a typically underutilized extremely high-performance resource: local NVMe storage. This built-in storage tier offers dramatically better cost-performance characteristics compared to any external storage solution available in the market.
Leveraging a standards-based approach using Linux pNFS v4.2 with Flex Files, Hammerspace enables organizations to activate this local NVMe as shared, protected Tier 0 storage, transforming this idle capacity into a seamless tier within a global namespace of multiple storage types from any vendor. Additionally, it provides automated data orchestration and data protection without requiring proprietary software.
As a result, longer context windows in AI inferencing can be sustained directly on local storage, significantly reducing the frequency of costly and latency-inducing data transfers to external storage. This approach utilizes existing infrastructure, eliminating the need for additional power or costly network upgrades, which substantially reduces both the cost and complexity of implementing AI projects.
In cloud environments, activating local NVMe within compute nodes similarly provides unmatched performance, dramatically outperforming traditional cloud storage bandwidth constraints and accelerating time-to-value for AI workloads.