The Real Bottleneck in High-Performance Computing
High-performance computing (HPC) environments need to operate at massive speed and scale to meet the demands of today’s complex workloads.
But even with the latest GPUs and optimized code, performance breaks down when the data infrastructure can’t keep up.
- Scratch storage fills up too fast, clogging the pipeline for new jobs
- Performance suffers when you can’t keep up with the speed of compute
- Old files linger, wasting high-value performance storage
- Manual data movement drains admin hours and introduce risk
- Data is spread across silos and systems, making it hard to locate what’s needed
- Rarely accessed data sits on expensive storage, inflating costs and reducing efficiency
HPC Storage That Finally Works Like It Should
HPC scratch storage is built for high-speed ingest, rapid checkpointing, and fast job output. But too often, it becomes a dumping ground for files that were never moved, archived, or cleaned up.
That’s where DataCore Nexus comes in—a unified data platform built to accelerate Tier-0 scratch and keep it clean and always ready.
It’s powered by two tightly integrated technologies
- Pixstor, for blazing-fast Tier-0 performance
- Ngenea, for intelligent data movement across every storage tier and global namespace access
Together, they ensure your compute stays fed, your workflows stay fluid, and your team stays focused on results.
Built to Match the Pace of Your Compute
Nexus brings together performance at the core and data orchestration across the lifecycle—all through a policy-driven platform that adapts to your workflows.
Performance-Driven Tier-0 Scratch
A powerful, software-defined parallel file system designed for the most demanding HPC workloads.
Key Features:
- NVMe-powered, low-latency performance
- Up to 180 GB/s read performance in a compact 2U form factor
- Seamless scale-out to petabytes and billions of files
- Resilient, highly available architecture
- No data lock-in or format tax
Benefits:
- Maximize compute efficiency—no more idle GPUs
- Minimize latency across parallel read/write workflows
- Avoid bottlenecks and I/O inconsistencies during HPC operations
- Future-proof HPC scaling without costly redesigns
Intelligent Data Movement Across All Tiers
Modern data management that turns your fragmented storage into a dynamic, policy-driven ecosystem.
Key Features:
- Automated tiering between scratch, home, archive, and cloud
- Global namespace for a single view of all unstructured data
- Policy-driven management to simplify data placement and recall
- Integration with HPC schedulers through REST APIs
Benefits:
- Keep scratch space lean—older data is offloaded automatically
- Recall datasets on access, without user intervention
- Align data placement with cost and performance needs
- Empower remote teams with global collaboration and instant data access
