Higher Education & Research

Academic research groups face a structural compute problem. Simulation-heavy disciplines — computational mechanics, quantum chemistry, climate modeling, bioinformatics — require HPC capacity that university data centers rarely deliver at the right time, right scale, or right cost. Shared cluster queues introduce unpredictable delays. Legacy infrastructure limits core counts and GPU availability. Procurement cycles that span semesters are incompatible with grant timelines. The result: researchers either wait, or pay cloud rates that quickly exceed project budgets for sustained workloads.

Mevasis provides academic teams with the missing middle ground: on-demand HPC access that scales with the research calendar, and procurement paths that fit grant-funded project structures.

Key Workloads

Research HPC usage spans a broad range of disciplines and software stacks. The table below covers the most common workloads and their infrastructure requirements.

Open-Source Software Stack

University research is predominantly built on open-source tools. Mevasis HPC environments are pre-configured for these workloads — no license overhead, no vendor lock-in:

• OpenFOAM: Full CFD pipeline, MPI-parallel, scales to hundreds of cores
• GROMACS / LAMMPS: GPU-accelerated molecular dynamics with CUDA/HIP support
• Quantum ESPRESSO: Plane-wave DFT; benefits strongly from InfiniBand at 64+ cores
• ORCA: Ab initio and DFT chemistry; high-memory nodes for large basis sets
• WRF: Weather Research and Forecasting; wide MPI scaling for regional climate runs
• MATLAB / Python: Interactive and batch modes via SLURM job scheduler

Application Areas

Research Group Compute Access

The most common pain point for academic HPC users is not availability in principle — it is timely availability in practice. Shared national or university clusters impose multi-day queue waits during peak periods (semester deadlines, conference submission windows). A research group with an active grant cannot afford to have simulations idle in a queue for 72 hours.

HPC Rental from Mevasis gives research groups dedicated, on-demand access sized to the project:

• Short-term bursts: Conference deadlines, grant deliverable periods, thesis crunch time
• Sustained project compute: Full-semester or full-year access with predictable monthly billing
• Grant-aligned billing: Usage can be tied directly to project cost codes; invoicing structured to match grant reporting periods

This removes the dependency on shared infrastructure and eliminates queue uncertainty entirely.

Student Competition Projects

Engineering student teams — Formula SAE, Formula Student, Baja SAE, SAE Aero Design, and similar competitions — require CFD and FEM simulation capability that is not available through standard university IT. Aerodynamics optimization, chassis structural analysis, and suspension kinematics all benefit from or require HPC.

Mevasis provides short-term rental access specifically sized for student team budgets: a 32–64 core CPU cluster with fast scratch storage is sufficient for most competition simulation workloads and can be provisioned and returned within a single academic term.

Interdisciplinary and Multi-Lab Projects

Large research projects funded under TÜBİTAK, EU Horizon, or bilateral programs often involve multiple research groups with heterogeneous compute needs. A climate modeling team and a materials simulation team in the same project have very different hardware profiles — but they share the need for a common, auditable compute environment.

Mevasis HPC Consulting supports multi-group cluster design: shared SLURM partition structures, per-project storage quotas, and reproducible software environments via containers (Singularity/Apptainer) or environment modules.

Reproducibility note: Containerized software stacks (Apptainer/Singularity) allow exact replication of the compute environment used to generate published results — directly supporting data management plan requirements for EU-funded projects.

Typical HPC Configuration

The following reference configuration covers a mid-size research group or small institute running mixed CPU and GPU workloads:

Login / Pre-/Post-Processing Nodes (2×)
├── CPU Compute Nodes (16–32 units)
│   └── 2× AMD EPYC 9654 (96 cores/node), 256–512 GB DDR5
│       (OpenFOAM, Quantum ESPRESSO, WRF, MATLAB batch)
├── GPU Compute Nodes (4–8 units)
│   └── 2× Intel Xeon + 4× NVIDIA L40S or A100
│       (GROMACS, LAMMPS, PyTorch, ORCA GPU)
├── High-Memory Nodes (2 units)
│   └── 2× EPYC, 1–2 TB DDR5
│       (large basis set quantum chemistry, big NGS analysis)
└── Scratch Storage
    └── BeeGFS on NVMe, 500 TB usable, 8+ GB/s aggregate

Scheduler: SLURM with fair-share and QOS partitions per group
Network: InfiniBand HDR100 (100 Gb/s) or HDR200, fat-tree

For student teams or individual research groups, a lighter configuration (8–16 CPU nodes, shared storage, Gigabit Ethernet) is sufficient and significantly reduces cost.

Mevasis Higher Education HPC Services

Mevasis works with research teams, department IT staff, and university procurement offices to match infrastructure to actual workload requirements — not to generalized specs.

• Workload profiling: Analyze your existing simulations to right-size hardware before procurement
• Flexible access models: Rental for project-duration access; on-premise installation for longer institutional commitments
• Software environment setup: Module system, Singularity containers, SLURM configuration for multi-group environments
• Student team support: Short-term competition-cycle rental with low entry cost
• HPC Rental: On-demand compute sized to grant budgets and academic calendars
• HPC Consulting: Architecture design, performance tuning, and reproducibility best practices

Frequently Asked Questions

How does HPC rental billing work for grant-funded projects? Mevasis offers monthly or project-duration billing with invoicing structured to match standard grant reporting periods. Usage can be attributed to specific project cost codes. Contact us to discuss the structure that fits your institution’s financial reporting requirements.

Is cloud (AWS, Azure) more practical for occasional academic workloads? For short bursts (under ~500 core-hours/month), cloud can be cost-effective. Beyond that threshold — and particularly for sustained workloads during an active project — on-premise or dedicated rental is substantially cheaper. The other advantage of dedicated rental: you are not competing with other tenants for network bandwidth, and InfiniBand MPI performance is consistent, not variable.

How are student competition projects handled? Is there a minimum commitment? There is no minimum term for student team rentals. A cluster can be provisioned for a single month and decommissioned after the competition deadline. Mevasis has supported Formula Student and SAE Aero Design teams at this scale.

Can Mevasis help set up a reproducible research computing environment? Yes. We configure Apptainer/Singularity container infrastructure alongside traditional environment modules, so research groups can archive the exact software environment used for a given publication. This directly satisfies data management plan requirements for TÜBİTAK and EU Horizon grants.