HPC for Universities: Building Academic Research Infrastructure Guide

Turkey has over 200 universities, with thousands of researchers working on compute-intensive simulations, big data analytics, and AI models. How many have their own university HPC infrastructure? Surprisingly few.

While centralized resources like TRUBA (Turkish National Science e-Infrastructure) and ULAKBIM are valuable, queue times are growing with increasing user numbers, and researchers are seeking institution-specific solutions. This guide covers the strategic, technical, and financial decisions a university faces when building its own academic research cluster.

Current State of Academic HPC in Turkey

Research university HPC deployments in Turkey have gained significant momentum over the last five years. Leading institutions like ITU, METU, Koc University, Sabanci University, and Bilkent operate their own clusters. However, many universities in Anatolia remain entirely dependent on TRUBA.

TRUBA’s current capacity as of 2026 is approximately 70,000 CPU cores and 1,000+ GPUs. While impressive in isolation, this is shared across 200+ universities and research centers. Wait times can reach weeks, especially for GPU-accelerated workloads (deep learning, molecular dynamics).

Universities with their own clusters share a common trait: institutional HPC is positioned not just as an IT investment, but as a cornerstone of research strategy. Researchers with dedicated infrastructure produce publications faster, can apply for larger projects, and provide graduate students with hands-on HPC experience.

Shared vs Departmental: Cluster Organization Models

The hybrid model is the most widely adopted approach today. A central cluster provides baseline compute capacity for all faculties, while additional resources such as a dedicated GPU cluster for engineering or high-memory nodes for science are positioned departmentally.

Multi-Tenant Configuration

A central cluster hosting multiple faculties and research groups requires well-designed multi-tenant architecture. SLURM offers flexible and powerful tools for this. See SLURM HPC Cluster Management for detailed coverage; here we highlight university-specific critical points.

Account and Project Structure: Each faculty or research group is defined as a SLURM account, with project-based sub-accounts underneath. This enables:

• Separate resource usage reporting per group
• Priority scheduling for funded projects (e.g., TUBITAK)
• Long-term balance through fairshare scoring

Quotas and Limits: Different faculties have different needs. Medical faculty requires high-memory nodes for genomics analysis, while engineering uses multi-core CPU nodes for CFD simulations. SLURM QOS (Quality of Service) definitions set group-specific resource limits.

Recommended initial configuration:

Account structure:
  ├── rectorate (administrative account)
  ├── science-faculty
  │   ├── physics-department
  │   └── chemistry-department
  ├── engineering-faculty
  │   ├── mechanical-department
  │   └── computer-science-department
  └── medical-faculty
      └── biostatistics

For more on multi-tenant configuration, see Multi-Tenant HPC Cluster.

Funding: TUBITAK, EU Grants, and Alternative Sources

The biggest barrier to a university HPC infrastructure is typically budget. Fortunately, various funding sources exist in Turkey and Europe.

TUBITAK Grants

• TUBITAK 1001 — Scientific and Technological Research Projects: HPC hardware can be requested under the “research infrastructure” budget item. Upper limit: 720,000 TL per project (2026).
• ARDEB — Infrastructure Support Programs: Special calls for large-scale research infrastructures. Projects over 5 million TL.
• ULAKBIM Collaboration: Expanding existing TRUBA infrastructure with institution-specific nodes.

EU and International Funding

• Horizon Europe — Research Infrastructures: Participation in European research infrastructure networks and shared resource usage.
• EuroHPC Joint Undertaking: Access to European supercomputing ecosystem and national competence center setup support.
• Erasmus+ Capacity Building: Joint infrastructure projects through inter-university collaboration.

Institutional Resources

• BAP (Scientific Research Projects): Most universities allocate infrastructure funds from their own budget. A well-justified feasibility study can secure meaningful budgets from BAP commissions. See HPC Investment Feasibility Guide for help preparing your feasibility report.
• Rental model: Budgeting as monthly/annual operational expense instead of large upfront capital. Especially compatible with BAP and revolving fund income. See HPC Rental for details.

Sample Architecture: Typical Cluster for a Mid-Scale Research University

A concrete reference cluster architecture suitable for a university with 10,000-20,000 students and active research output:

Estimated budget: $600,000 – $900,000 USD (hardware, installation, 3-year support). Under a rental model, this drops to $18,000 – $27,000 USD/month.

Installation time: 6-8 weeks after site preparation (hardware procurement + physical installation + software configuration + acceptance testing).

For detailed roles and specifications of each component, see HPC Cluster Components. Our HPC Consulting service covers the full process from needs analysis to acceptance testing.

Next Step with Mevasis

If you are considering an academic research cluster for your university, Mevasis supports you at every stage:

1. Needs Analysis: Map your current research workloads and determine which hardware will serve which faculty.
2. Architecture Design and Budgeting: Prepare a technical specification aligned with your university scale, budget, and TUBITAK/BAP funding programs.
3. Turnkey Installation: Manage the entire technical process from site preparation to acceptance testing.
4. Training and Handover: Teach your system administrators and researchers to use the cluster efficiently through HPC Training programs.

Review our completed university projects on the Case Studies page. Contact us for a custom preliminary assessment.

Frequently Asked Questions

What is the minimum HPC budget for a small university?

A small cluster with 8-16 CPU nodes and 1-2 GPU nodes can start from $5,000-8,000/month under a rental model. For purchase, an initial budget of $150,000-250,000 USD is sufficient. This scale meets the basic needs of 3-5 research groups.

Is TRUBA not sufficient?

TRUBA is a shared national resource, and wait times can be long, especially for GPU workloads. Additionally, institution-specific software licenses (ANSYS, MATLAB, Gaussian, etc.) may not always be usable on TRUBA. Your own cluster provides priority access, custom configuration, and license flexibility.

How long does TUBITAK funding take?

TUBITAK 1001 applications are evaluated twice per year. The time from application to fund availability is typically 8-12 months. EU funding can take 12-18 months. For rapid startup, the rental model provides infrastructure access without waiting for the funding process.

How long does installation take?

After site preparation (power, cooling, network infrastructure), hardware installation and software configuration take 6-8 weeks. Mevasis manages this process turnkey, delivering the system ready for use including acceptance testing.