HPC buyers already speak stack-native language — InfiniBand, MPI, Lustre, SLURM. The GLabs HPC Core is that stack as one unified view: four layers, each with its own reference hardware and software, instead of four separate pages.
Deployed with GLIDS — the GLabs Intelligent Deployment System.
GLabs HPC Core
| Layer | What It Covers | Reference Hardware / Software |
|---|---|---|
| Compute | CPU-bound MPI through GPU-accelerated and mixed AI/HPC node types. | General Purpose & Multi-node chassis for CPU-bound; NVIDIA MGX platforms for GPU-accelerated. |
| Fabric | RoCEv2 100GbE up to InfiniBand NDR 400G, matched to cluster tier. | Selected per deployment tier — full tiers on GLabs Enterprise HPC. |
| Storage | Lustre / BeeGFS parallel filesystem, with a cold tier to Vault Infinite. | GLabs Enterprise Vault & POD (120GB/s NVMe-oF, GPU Direct Storage). |
| Integration | SLURM, Singularity/Apptainer, OpenMPI, Grafana + Prometheus + DCGM, pre-pulled scientific containers. | Software layer — platform-agnostic across the Compute tier above. |
| Cluster Tier | Nodes | Fabric |
|---|---|---|
| HPC Edge | 4–16 | 100GbE |
| HPC Research | 16–64 | InfiniBand HDR 200G |
| HPC Enterprise | 64–256 | InfiniBand NDR 400G |
| HPC Sovereign | 256–2,000+ | InfiniBand NDR 400G / Quantum-X800 |
From EuroHPC to Owned Compute — teams migrating off shared EuroHPC allocations or national compute grants land directly on an owned HPC Enterprise or HPC Sovereign tier, with the same SLURM/Lustre environment they already run against.
Use Cases
HPC environments sized for molecular dynamics and structural biology workloads — protein folding, docking simulations, cryo-EM reconstruction — where the job either fits the cluster's memory and interconnect profile or it doesn't run at all.
Built On: GROMACS, NAMD, LAMMPS, AlphaFold3, CryoSPARC, RELION.
Engineering simulation at a scale that outgrows a workstation — full-vehicle CFD, multi-million-cell meshes, coupled structural-thermal FEA — distributed across cluster nodes rather than queued on a single machine.
Built On: ANSYS Fluent, OpenFOAM, STAR-CCM+, Abaqus.
Large-scale climate, weather, seismic, and subsurface modelling workloads that run across hundreds of nodes for days at a time, where fabric latency and parallel storage throughput matter as much as raw core count.
Built On: WRF, MOM6, OpenIFS, SPECFEM3D, Petrel.
Computational chemistry and materials-science applications — electronic structure calculations, materials property prediction — that are as memory-bandwidth-bound as they are compute-bound.
Built On: CP2K, ORCA, VASP, PSI4, Quantum ESPRESSO.
ExploreGLabs Enterprise HPCfor full node profiles, fabric tiers, and reference hardware.
A solutions architect responds directly — no ticket queue.
Talk to a Solutions Architect