- REV
- 2026.07
- BUILD
- 05d82796 · 24 JUL
- STACK
- NEXT 15.5 · STATIC EXPORT
Jean-Sébastien Loiseau
Infrastructure & Systems Engineer
Research & Infrastructure Engineer at the University of Reims. Working my way into HPC, node by node.
VIRTUALIZATION · LINUX · NETWORKING · SELF‑HOSTED PLATFORMS
- REV
- 2026.07
- BUILD
- 05d82796 · 24 JUL
- STACK
- NEXT 15.5 · STATIC EXPORT
Served from my own rack. Static export: no tracker, no cookies.
The isometric datacenter is an illustration; its labels are scenery. The real infrastructure is one self-hosted rack serving this site.
Click a rack to open it.
HPC-LABPROVISIONING
LEARNING QUEUE · PROVISIONINGI am heading toward high performance computing and this rack says so honestly: it is being provisioned. The list below is my real roadmap, and its milestones go LIVE as I do. Target playground: LANL's public Darwin cluster (CCS-7).
| ID | NAME | GOAL | STATE |
|---|---|---|---|
| 001 | infra-foundations | Virtualization, networking, automation and observability: the infrastructure layer every HPC platform sits on. | COMPLETED |
| 002 | hpc-hardware | Installed H100 GPUs in new HP servers; took part in the architecture discussions on exposing them (PCI passthrough vs vGPU, Nvidia licensing constraints, and why a container platform simplifies GPU operations compared to VM-based passthrough). | COMPLETED |
| 003 | slurm-basics | Stand up a mini Slurm cluster: controller, login node, and four compute VMs across two partitions, on Proxmox. | LIVE |
| 004 | lustre-storage | Understand and test parallel storage. | PLANNED |
| 005 | mpi-hello | First MPI programs on the home cluster. | PLANNED |
| 006 | cluster-lab | Benchmark and document a small compute lab. | PLANNED |
001infra-foundationsCOMPLETED
Virtualization, networking, automation and observability: the infrastructure layer every HPC platform sits on.
002hpc-hardwareCOMPLETED
Installed H100 GPUs in new HP servers; took part in the architecture discussions on exposing them (PCI passthrough vs vGPU, Nvidia licensing constraints, and why a container platform simplifies GPU operations compared to VM-based passthrough).
003slurm-basicsLIVE
Stand up a mini Slurm cluster: controller, login node, and four compute VMs across two partitions, on Proxmox.
004lustre-storagePLANNED
Understand and test parallel storage.
005mpi-helloPLANNED
First MPI programs on the home cluster.
006cluster-labPLANNED
Benchmark and document a small compute lab.
RACK PROVISIONING2 COMPLETED1 LIVE3 PLANNED
VIEW 1 / 4