See Tier 0 Checkpointing Live
The fastest storage in your cluster is the NVMe already inside your GPU servers. Watch it become shared, protected, checkpoint-ready storage in a single global file system. No new hardware, no application changes.
What the demo covers:
What the demo covers:
- Activating server-local NVMe as Tier 0: shared across nodes, protected, in the global namespace.
- Checkpointing and restart at PCI bus speed, with automated tiering to scratch and archive behind the scenes.
- Standard NFSv4.2 and pNFS access from unmodified Linux clients: no proprietary client software, nothing to install on compute nodes.
Tier 0 means local SSD inside the compute cluster, integrated into a global namespace and orchestrated to all other tiers. Other vendors use the term loosely. This demo shows the rigorous version.