It looks like #Excelero has re-emerged after being acquired by #NVIDIA! A partially hyperconverged #Lustre-on-Excelero-on-DGX is the high-performance file solution for NVIDIA DGX Cloud in #OCI.

It looks a bit complicated; parts of Lustre are running on DGX, and other parts run on HA pairs of storage nodes. No mention of the data management or resilience models though. Wonder what happens when a DGX drops a GPU and the node fails.

https://developer.nvidia.com/blog/high-performance-storage-on-nvidia-dgx-cloud-with-oracle-cloud-infrastructure/

#HPC #AI #storage