ParTec

Technology
MODULARITY – Proven at Exascale. Ready for Post-Exascale. Built to Evolve.

For a sovereign and energy-efficient digital future,
modularity at the system level is not optional — it is a key requirement.

For a sovereign and energy-efficient digital future, modularity at the system level is not optional — it is a key requirement.

Modern HPC, AI, and Quantum computation demands heterogeneity. Our architecture and orchestration software ensure adaptability and operate as a unified system, eliminating technological dead ends, preventing vendor lock-in, and enabling controlled technological evolution with long-term investment security.

With ParTec’s patented dynamic Modular System Architecture (dMSA) we create a unique structural nexus between diverse compute elements — CPUs, GPUs, accelerators, storage, memory, and future quantum technologies — composing them into one adaptable, energy-aware system. 

dMSA underpins Europe’s most powerful and sustainable supercomputers.

Our Technological Advantages

Architecture

Modularity as the Architectural Foundation of Sovereign Compute

The Backbone of Europe’s first Exascale System and other leading Supercomputers

The underlying system infrastructure determines long-term capability, investment security, technological independence, and innovation potential. A well-designed architecture enables systems to evolve over time rather than reaching structural limits.

Architecture and orchestration ultimately determine whether infrastructure can continuously adapt to new requirements.

Only scalable, modular architectures enable true sovereignty. The architectural choices made today determine whether infrastructure remains adaptable — or becomes obsolete.

The Architectural Challenge of Conventional Architectures

Conventional supercomputers are built as fixed systems in a world where workloads are nowadays highly dynamic.

The consequences are measurable. This rigidity leads to 

  • Underutilized resources,
  • Higher energy consumption,
  • Rising costs,
  • Reduced competitiveness,
  • and long-term vendor lock-in.
Homogenous accelerated clusters replicate identical node designs. Each node combines CPUs and accelerators in a fixed ratio. This model was sufficient when workloads were predictable and had uniform computational requirements. This is no longer the case for modern computing infrastructures.
 

Computing environments now need to accommodate a wide swathe of workloads, which include:

  • Simulation
  • AI training and inference
  • Data analytics
  • Hybrid HPC–AI workflows
  • Emerging quantum-classical models
  • … and more!

Fixed-node architectures force heterogeneous workloads onto homogeneous hardware, making compromises which impact delivered performance and efficiency. This constraint is structural. 

The Architectural Solution: Modularity

ParTec’s dynamic Modular System Architecture (dMSA)

ParTec’s patented dynamic Modular System Architecture (dMSA) enables scalable, hardware-agnostic, and energy-efficient heterogeneous computing for HPC, AI and QC — by design. 

Proven at Scale

Based on extensive research and many years of expertise, ParTec’s dMSA forms  the structural backbone of Europe’s leading high-performance systems.

The dMSA has been deployed in leading European supercomputers and serves as the architectural blueprint for European Exascale and post-Exascale systems.

Future-Ready by Structure

The architecture anticipates technological evolution.

Quantum processors, storage, memory, and emerging compute paradigms such as neuromorphic computing can be integrated as additional modules and scaled independently— without dismantling the system foundation. There are no technological dead ends – only controlled evolution, flexibility, and scalability exactly as the demands require.

Customization by Design

One of ParTec’s core strengths is co-design and deep customization.

Through our modular architecture, systems are precisely aligned with specific workload characteristics, energy profiles, and long-term strategic objectives. Infrastructure is not predefined. It is engineered to match current demand — and structured to evolve with future requirements. Hence, our customers can benefit from innovations even long after the deployment of their systems.

Architecture in Action

The dMSA replaces node replication with tightly coupled compute and storage modules.

Each module is purpose-built:

  • Accelerator modules for massively parallel execution
  • CPU modules for serial and latency-sensitive workloads
  • AI-optimized configurations
  • Dedicated storage and memory tiers

Applications are no longer constrained by fixed ratios of different hardware entities. Runtime systems dynamically compose execution environments across modules — matching architecture to workload characteristics in real time. Composition and decomposition occur during runtime. Different execution phases operate on different architectural configurations.
The system adapts structurally.

The dMSA enables:

  • Precise workload-to-resource alignment
  • Improved time-to-solution
  • Reduced energy-to-solution
  • Higher throughput and system utilization
  • Scalable heterogeneity
  • Lifecycle adaptability without architectural redesign

Highly parallel components run on dense accelerator modules. Serial or control-intensive phases run on high-performance CPUs. Efficiency is achieved by design.

Mixed HPC–AI workflows benefit particularly from this structural flexibility. 

SOFTWARE

ParaStation Modulo Software Suite

The Backbone for Sovereign Modular Compute

If hardware is the body of a supercomputer,
architecture is its structural design —
and orchestration is its nervous system.

ParaStation Modulo Software Suite functions as the backbone of the supercomputer,— the governing layer that coordinates CPUs, GPUs, accelerators, storage, and emerging technologies into one coherent operational fabric.

Without this layer, infrastructure remains fragmented and inefficient.

With it, infrastructure becomes not only adaptive, governable, and resilient but also highly flexible and scalable, ready to evolve with true workload demands.

From Architecture to Operational Sovereignty

Complementing our dynamic Modular System Architecture (dMSA), the ParaStation Modulo Software Suite delivers:

  • Workload flexibility
  • Reliability due to reduced system downtime and failures, maximizing system utilization for uninterrupted performance
  • Technological scalability
  • Energy-aware system optimization
  • Hardware-agnostic integration of customer desired modules such as CPU, GPU, storage, memory, QC, and other disruptive technologies

These capabilities translate into:

  • Reduced total cost of ownership
  • Enhanced long-term investment security
  • Sovereign control over infrastructure evolution
  • Lower energy-consumption
  • Maximized system utilization under dynamic workload conditions

The Strategic Impact of the ParaStation Modulo Software Suite

Architectural Sovereignty

The ParaStation Modulo Software Suite eliminates structural vendor lock-in through hardware-agnostic integration and modular orchestration. Emerging technologies integrate without redesigning the system foundation.

Control over long-term compute capability remains with the operator — not the vendor.

Lifecycle Investment Protection

Infrastructure decisions define many years of capability and innovation.

The ParaStation Modulo Software Suite enables controlled system evolution, providing high protection of investments from technological obsolescence. Modular integration avoids disruptive redesign and enables expansion without architectural replacement.

Systemic Orchestration of Heterogeneity

Modern environments combine HPC-style simulation, AI, analytics, and emerging paradigms within one infrastructure.

The ParaStation Modulo Software Suite dynamically composes heterogeneous modules into unified execution environments, aligning resources with workload requirements in real time across the system fabric.

Heterogeneity becomes scalable, performance-optimized, and economically efficient.

Energy-Performance Optimization

Through system-level workload governance and architectural alignment, the ParaStation Modulo Software Suite enables performance-per-watt optimization across heterogeneous modules.

Operational Continuity at Scale

The ParaStation Modulo Software Suite is built for production environments at the highest scale. Integrated monitoring, validation, lifecycle management, and proactive support ensure resilience and predictability.

Exascale-Proven Performance and Efficiency

The ParaStation Modulo Software Suite serves as the orchestration backbone of Europe’s first Exascale system — and the most energy-efficient Exascale system worldwide.

Core Components

The ParaStation Modulo  Software Suite is a comprehensive system management and orchestration framework consisting of four tightly integrated components:

ParaStation ClusterTools

Hardware-agnostic provisioning and administration framework simplifying deployment and lifecycle management. An image-based node management approach allows for rolling updates and minimal downtime. It integrates seamlessly with tools for efficient configuration management.

Scalability & Cluster Management

Managing large clusters with heterogeneous components and minimizing downtime is complex. System admins need reliable ways to provision, update, customize thousands of nodes at scale.

ParaStation HealthChecker

Automated integrity checks with non-destructive testing, reporting, and a self-recovering “checkbot” that brings offline nodes back online whenever possible.

Ensuring System Health

Component failures, system errors and misconfigurations can compromise productivity and cause significant delays in job execution, especially in large-scale, modular systems.

ParaStation TicketSuite

Thanks to seamless integration with ClusterTools and HealthChecker, TicketSuite centralizes all issue tracking and escalation in one interface dedicated to the system at the customer’s site.

Efficient Issue Tracking

Managing and resolving system-level issues across multiple modules and nodes can be overwhelming, particularly in distributed or hybrid environments.

ParaStation MPI

A Message-Passing Interface (MPI) runtime library with optimized low-level, point-to-point communication and support for the PMIx standard, enables high performance for malleable workloads and evolving workflows.

Communication For Dynamic Workloads

Low-latency, high-bandwidth communication across modular, heterogeneous systems is essential for Exascale computing but is challenging in federated networks and for highly dynamic workloads.