HPC Infrastructure Deployment — Turnkey Cluster Delivery

HPC infrastructure deployment means running every stage — from design decisions through physical installation and software configuration to user handover — with a specialist team. Mevasis delivers enterprise HPC projects for customers ranging from universities and pharmaceutical companies to defence organisations and the finance sector.

Deployment Methodology

Phase 1: Design and Procurement (6–14 weeks)

Technical design:

• Workload analysis and sizing
• Node type and count
• Network topology (InfiniBand fat-tree, Ethernet)
• Storage architecture
• Power and cooling calculation

Procurement:

• Hardware ordered from approved suppliers
• Delivery tracking and logistics coordination
• Pre-acceptance testing at the factory

Phase 2: Physical Installation (1–2 weeks)

• Rack layout plan (1U/2U node placement, switch position)
• Cable management (Ethernet, InfiniBand, power)
• PDU installation and power balancing
• Out-of-band (OOB) management network
• Cooling airflow optimisation

Phase 3: Software Stack (2–3 weeks)

1. Operating system (Rocky Linux 9)
   - PXE boot infrastructure
   - Warewulf 4 cluster provisioning
   - Node image creation and distribution

2. Network configuration
   - InfiniBand subnet manager (OpenSM)
   - IP addressing (management, compute, storage)
   - DNS and NTP

3. Job scheduler
   - SLURM installation and configuration
   - MUNGE authentication
   - Accounting database (slurmdbd)
   - Partition and QOS definitions

4. Software environment
   - Lmod module system
   - Compilers (GCC, Intel oneAPI)
   - MPI libraries
   - CUDA + cuDNN (for GPU nodes)

5. Monitoring
   - Prometheus + Grafana
   - DCGM Exporter
   - Alertmanager

Phase 4: Testing and Acceptance (1–2 weeks)

• HPL (Linpack) benchmark: targeting 75%+ efficiency against theoretical peak
• IOR storage benchmark: I/O throughput verification
• OSU MPI benchmark: latency and bandwidth measurement
• GPU benchmark: nvidia-smi and DCGM health checks
• User acceptance testing: validation against real workloads

Phase 5: Handover and Training

• System administrator training (2 days)
• User training (SLURM, Lmod, best practice)
• Documentation handover
• Start of 90-day warranty support

To get a quote for HPC infrastructure deployment, get in touch.

Hardware Catalogue

You can review the network hardware we deploy in our HPC networking catalogue and request a quote for a part list matched to your requirement.