HPC Migration — Moving Existing Infrastructure to a Modern System

HPC infrastructure ages: operating system support ends, scheduler technology moves on, hardware capacity falls short. A migration project needs expert planning to modernise the system while production workloads keep running.

Typical Migration Scenarios

CentOS 7 → Rocky Linux 9 / AlmaLinux 9

CentOS 7 reached end of life in June 2024. An EOL system receives no security patches, which is a serious risk for enterprise infrastructure.

The Mevasis migration process:

1. Inventory of current packages and dependencies
2. Rocky Linux 9 test environment build
3. Application compatibility testing
4. Staged production cutover, node by node
5. Validation and monitoring

SGE / PBS → SLURM

SLURM has become the industry standard. Moving from SGE (Son of Grid Engine) or older PBS releases requires job script compatibility work.

# SGE job script → SLURM equivalent
# SGE:   #$ -pe mpi 32
# SLURM: #SBATCH --ntasks=32

# SGE:   qsub my_job.sh
# SLURM: sbatch my_job.sh

Cloud → On-Premises (Repatriation)

Moving back to owned infrastructure because of a high cloud bill or data security requirements. Mevasis analyses the workload profile to determine the right on-premises sizing.

Hardware Refresh

Replacing an existing cluster with new hardware while keeping workloads running. Parallel environment build, data movement and staged user migration.

Migration Methodology

Discovery: current system inventory, dependency map, user analysis

Test environment: a test cluster that mirrors production

Pilot: trial runs on the new system with selected workloads

Parallel run: both systems active, users move across in stages

Cutover: legacy system retired, new system fully in production

Watch period: 2–4 weeks of close monitoring with rollback ready

To scope your migration project, arrange an assessment call.