Government & Defense

Defense industry and public sector organizations operate under constraints that civilian compute environments are not designed to meet. Data classification requirements, export control regulations, and institutional security policies collectively rule out public cloud as a platform for sensitive simulation workloads. At the same time, the compute demands of modern defense R&D — radar cross-section modeling, ballistic trajectory analysis, signal processing, electronic warfare simulation — are substantial and growing.

On-premise HPC infrastructure, purpose-built for classified or export-controlled environments, is the only architecture that satisfies both the technical and the regulatory requirements simultaneously. Mevasis designs, procures, and deploys these systems.

Regulatory Framework: What Requires On-Premise Infrastructure

The most important question for defense and government organizations is not “should we use HPC?” but “where must we run it?”

ITAR and EAR Export Control

The U.S. International Traffic in Arms Regulations (ITAR) and Export Administration Regulations (EAR) impose strict controls on defense-related technical data. For Turkish defense industry companies working under co-production, licensed manufacture, or joint development agreements, ITAR/EAR compliance obligations frequently extend to the computing infrastructure used to process that data.

Workloads that typically require on-premise infrastructure under export control obligations:

• Ballistic trajectory and terminal effects simulation
• Radar cross-section (RCS) analysis using ITAR-controlled geometry or signature data
• Electronic warfare system performance modeling
• Propulsion system simulation (missile, rocket motor)
• Ordnance and warhead design analysis

Cloud providers, even those with government-specific offerings, do not universally satisfy the access control and data residency requirements that ITAR-controlled data demands in the Turkish regulatory context. On-premise infrastructure, within the organization’s physical security perimeter, is the definitive solution.

Data Classification and Access Control

Defense organizations operate with multiple classification tiers, each requiring physical and logical separation. HPC infrastructure must be designed from the ground up to enforce these boundaries — not retrofitted to approximate them.

Mevasis implements segmented cluster architectures with:

• Physical network separation between classification levels (no shared switching fabric)
• Separate storage namespaces per classification tier, with independent access control
• Audit logging of all login events, job submissions, and data transfers — tamper-evident, off-system log forwarding
• Role-based access control (RBAC) via LDAP/Active Directory integration with SLURM

Defense Industry Workloads

Ballistics and Terminal Effects

Ballistic simulation spans interior, exterior, and terminal phases. Each involves distinct physics and software:

• Interior ballistics: Combustion, propellant gas dynamics, projectile acceleration (LS-DYNA, AUTODYN, ALE3D)
• Exterior ballistics: 6-DOF trajectory modeling, aerodynamic coefficient derivation from CFD
• Terminal effects: Penetration mechanics, fragmentation, shaped charge jet formation (Abaqus explicit, LS-DYNA SPH/ALE)

These workloads are computationally intensive. A high-fidelity shaped charge detonation simulation using Smoothed Particle Hydrodynamics (SPH) can run on 64–256 cores for 12–48 hours per case. Parametric studies across detonation geometries require job arrays at scale.

Radar Cross-Section and Electromagnetic Simulation

RCS computation for aircraft, missiles, and surface platforms involves full-wave electromagnetic solvers across frequency bands. Key software includes:

High-frequency full-wave solutions for electrically large platforms (aircraft, ship) are among the most compute-intensive electromagnetic workloads. MLFMM (Multilevel Fast Multipole Method) formulations allow tractable solution times with adequate core counts, but memory requirements can reach 512 GB–2 TB per simulation.

Signal Processing and Electronic Warfare Simulation

EW system performance modeling — jamming effectiveness, intercept range prediction, radar detection probability — involves computationally intensive Monte Carlo simulations across scenario parameter spaces. These workloads are embarrassingly parallel: thousands of independent scenarios run simultaneously across CPU cores without inter-node communication.

SLURM job arrays on a 256–1,024 core cluster can process a full EW engagement scenario library in hours rather than days.

Radar and Sensor System Development

Phased array antenna pattern synthesis, SAR (Synthetic Aperture Radar) image formation, and sensor fusion algorithm development all involve significant numerical computation. GPU-accelerated signal processing (CUDA, cuFFT, cuBLAS) is the standard approach for large array simulations and SAR backprojection.

Public Sector and Defense Research Institutions

Beyond prime defense contractors, the following organization types have recurring HPC requirements:

• Defense research institutes and TÜBİTAK-affiliated centers: Basic and applied research with security obligations
• University defense research programs: Dual-use technologies, NATO-program participation
• Ministry-level R&D departments: National capability development, procurement support simulation

For these organizations, the procurement and operational model differs from commercial customers. Mevasis works within public tender frameworks and provides documentation appropriate for TÜBİTAK and government procurement processes.

Air-Gapped Cluster Architecture

For the highest-security workloads, air-gapped cluster deployment — physically isolated from all external networks — is the required architecture. Mevasis has designed and deployed air-gapped HPC systems with the following characteristics:

• No external network connectivity: Cluster operates on isolated private network with no gateway to internet or corporate WAN
• Controlled data ingress/egress: Formal data transfer procedures via approved portable media or one-way data diodes
• Physical access control integration: Server room access logging integrated with cluster audit trail
• Offline software management: Patch and update workflows that do not require internet connectivity on the cluster

Air-Gapped Cluster Reference Architecture
──────────────────────────────────────────
[Secure Perimeter / Server Room]

  Login Nodes (2×, HA)
  ├── CPU Compute Nodes (16–64 units)
  │   └── 2× AMD EPYC 9654, 512 GB DDR5
  ├── GPU Nodes (4–8 units, if required)
  │   └── NVIDIA A100 or L40S
  │       (RCS GPU solvers, EW Monte Carlo)
  ├── High-Memory Nodes (2–4 units)
  │   └── 1–2 TB DDR5
  │       (large EM models, SPH terminal effects)
  └── Storage (fully on-premise)
      ├── BeeGFS NVMe scratch: 100+ TB, 20+ GB/s
      └── Archive tier: encrypted SAS/NL-SAS, 500+ TB

Network:
  ├── InfiniBand NDR (compute fabric)
  └── Isolated management network (IPMI/BMC only)

No uplinks to external networks.
Data transfer: dedicated air-gap workstation with
  controlled media procedures.

Support Model

Defense and government systems cannot tolerate unresolved hardware failures. A compute node failure during a deadline-driven analysis program has direct schedule consequences.

Mevasis provides a support model designed for these requirements:

• 24/7 on-call technical support with defined escalation paths
• Defined SLA for critical fault response: hardware replacement and cluster recovery timelines agreed before deployment
• Spare parts inventory: Critical components (compute nodes, network switches, storage controllers) pre-staged or held in agreed reserve
• Scheduled maintenance windows: Coordinated with operational calendar to minimize impact on active programs
• On-site support availability: Response in person when remote resolution is insufficient

Mevasis Government & Defense HPC Services

Mevasis provides end-to-end HPC infrastructure services for defense industry and public sector organizations, from initial requirements analysis through long-term operational support.

• Requirements and compliance analysis: Identify which workloads require on-premise infrastructure and what security architecture is needed
• Air-gapped and secure cluster design: Physical and logical security architecture aligned to classification requirements
• Procurement support: Hardware specification, vendor management, and documentation for public tender processes
• Installation and commissioning: On-site hardware installation, network configuration, scheduler and security software setup
• Operational support: HPC Consulting and managed support with SLA guarantees appropriate for defense program timelines

Frequently Asked Questions

Which of our simulation workloads require on-premise infrastructure under ITAR? Any workload that processes ITAR-controlled technical data — geometry, performance parameters, design specifications — associated with USML-listed defense articles requires on-premise infrastructure. In practice, this includes most ballistic, RCS, and propulsion simulation work for programs with U.S.-origin technology content. Mevasis does not provide legal advice, but our technical team can work alongside your compliance office to map workloads to infrastructure requirements.

How is an air-gapped cluster managed without internet access? Offline software management uses pre-staged package repositories mirrored on an isolated internal server. Operating system patches, scheduler updates, and application software are tested on a separate validation system before being introduced to the air-gapped cluster via controlled procedures. This adds management overhead but is achievable with the right operational workflow, which Mevasis documents as part of the deployment.

What is your SLA for hardware failure response? SLA terms are agreed per contract. Typical defense program commitments include: 4-hour remote response for critical faults, next-business-day on-site response, and 48-hour hardware replacement for failed compute nodes from pre-staged spares. More aggressive SLAs are available for programs where cluster downtime has direct schedule impact.

Can you support procurement through public tender processes? Yes. Mevasis has experience providing technical specifications, compliance documentation, and pricing structures suitable for public tender submissions under Turkish procurement law. Contact us to discuss how HPC Consulting support can be scoped within a tender framework.