I will set up a linux hpc cluster with openhpc, slurm, and infiniband

Setting up an HPC cluster correctly from the start saves months of debugging and prevents the configuration debt that causes 80% of cluster performance problems later.

I've commissioned HPC clusters from 4 nodes to 600 nodes under India's National Supercomputing Mission from bare metal to HPL acceptance.

Full stack I work with:

Provisioning: Warewulf 4, xCAT, PXE  

OS: Rocky Linux 8/9, AlmaLinux, CentOS Stream  

Scheduler: Slurm with full accounting and cgroup  

MPI: OpenMPI, IntelMPI, MVAPICH2  

Fabric: InfiniBand HDR/NDR/EDR, Ethernet RDMA  

Storage: Lustre, BeeGFS, GPFS, NFS  

Monitoring: Grafana, Prometheus, Ganglia  

What you receive:

Fully provisioned compute nodes

Working Slurm queue with test jobs verified  

InfiniBand fabric validated with ibdiagnet

MPI hello world and bandwidth test passing

Complete configuration documentation

Handover call to walk you through the system

Before ordering: message me your node count, hardware specs, and what workloads you plan to run. I'll confirm feasibility and timeline before you pay.