ThinkParQ Launches BeeOND Enterprise: The On-Demand HPC Storage Solution That Turns Idle NVMe into a High-Performance File System

ThinkParQ, the company behind BeeGFS, announced the launch of BeeOND Enterprise (BeeGFS On Demand), a standalone subscription-based edition that gives AI and HPC teams the complete on-demand storage pipeline: local NVMe across compute nodes dynamically pooled into a high-performance scratch file system, spun up ins at job start and torn down at job end, plus BeeGFS Copy, the parallel data staging engine that stages input data into the scratch tier before a job begins and writes results back to global storage on completion, ensuring BeeOND’s NVMe speed benefits the full job pipeline, not just the computational phase.The launch addresses one of the most overlooked inefficiencies in high-performance computing: the terabytes of fast local NVMe storage sitting idle in every compute node while organizations continue to purchase additional dedicated burst-buffer appliances to address storage bottlenecks.

“The storage bottleneck in AI and HPC isn’t a hardware problem – the NVMe is already there, in every compute node, already paid for. BeeOND simply puts it to work,”  Frank Herold, CEO, ThinkParQ.

BeeGFS Copy: End-to-End Data Staging, Exclusive to Enterprise Edition
Fast scratch storage is only half of the pipeline. BeeOND Enterprise ships with BeeGFS Copy, a parallel data staging engine that moves data at scale between multiple persistent data lakes and the BeeOND scratch tier. It stages input data into the scratch tier before a job begins and writes results back to global storage on completion, ensuring BeeOND’s NVMe speed benefits the full job pipeline. 

Benchmarks validate BeeOND’s horizontal scaling claims with measured data, which demonstrated the utilization of up to 97% of the physical bandwidth of the aggregated  NVMe’s theoretical throughput. The results confirm that BeeOND’s performance scales predictably with node count, giving teams a reliable basis for capacity and IO/s planning: adding nodes to a job directly and proportionally increases available storage throughput.

BeeOND is operational across hundreds of organizations, from national laboratories to enterprise AI teams, at scales exceeding 10,000 compute nodes. 
BeeOND Enterprise is available, empowering every organization to accelerate results and maximize productivity, with no additional hardware investment.

ThinkParQ is showcasing BeeOND Enterprise at ISC High Performance 2026 in Hamburg at booth L40.