High-performance, scalable file and object storage for AI/HPC that improves flash performance, effective capacity, and endurance
Benefits
- Better system and application performance: Hammerspace efficient architecture combined with ScaleFlux NVMe SSDs deliver sustained, linearly scalable performance
- Faster query and transaction processing: The combined solution delivers fast responsiveness, high throughput, and low-latency across mixed read/write workloads
- Less Wasted Capacity and Lower Infrastructure Costs: Hammerspace parallel architecture cuts server and networking requirements by 2x vs scale-out NAS, while ScaleFlux write-reduction technology enables better capacity utilization and higher fill rates.
- Seamlessly Orchestrate Data Across Sites and Clouds: Hammerspace automates data movement in a global namespace that spans sites, clouds, and storage, to bring data to cloud compute resources
Certified Solution for High-Performance Flash Storage with Better Performance, Capacity, and Reliability
Today’s unstructured data workloads, including AI training and inference, require a data platform that delivers high-throughput, low-latency performance at large scale.
Hammerspace is a file and object data platform for hybrid- and multi cloud AI/HPC infrastructure. It is software-defined and storage agnostic, and seamlessly orchestrates data between sites, clouds, and any storage system.
ScaleFlux provides NVMe SSDs with unique System-on-Chip (SoC) controllers that improve flash performance, effective capacity, and endurance relative to ordinary enterprise NVMe SSDs.
Together, Hammerspace and ScaleFlux provide highly-performant, scalable, and reliable all-flash data storage for demanding file and object workloads.
Combined Solution Delivers Record-Setting Performance Results for AI/ML Workloads
Hammerspace uses ScaleFlux NVMe SSDs in its performance lab. ScaleFlux CSD5000 drives are used as the storage inside Hammerspace metadata servers (to improve metadata operations speed, effective capacity, and endurance), and in Hammerspace data servers that serve the data to clients (for example GPU nodes). This combined solution demonstrated record-setting MLCommons MLPerf Storage Benchmark results, exceeding the results of other leading HPC file systems.
ScaleFlux NVMe SSDs combine industry-standard form factors and multiple NAND options with innovative write-reduction technology and metadata management on their SoC controllers. The result is a multiplier in SSD efficiency – improved performance, capacity per $, and endurance relative to ordinary NVMe SSDs.
