Epyc
“The support from Qarnot should be an example to other companies. The way they approach partnership is extraordinary. They give us reliable compute power and a level of collaboration we simply did not get elsewhere. It feels less like working with a supplier and more like moving forward together as true partners.”
Freek Bos · IT Manager
Gauss Fusion aims to build a commercial fusion power plant in Europe using stellarator technology (magnetic confinement devices designed to stabilize plasma for fusion). The development of this plant includes multi-objective physics optimization, physics studies, engineering analysis, and the simulation work required to support these activities. The stellarator design and optimization tools are large physics codes designed to run on multi-node supercomputer architectures. Design development, validation, and safety studies, as well as geometry studies and physics calculations, rely on large-scale, complex simulations that require significant compute capacity.
Gauss’s team is facing increasingly demanding simulations, with heavy workloads, long runtimes, and ever-larger models. As parallel studies multiply and models grow in complexity, the need for computing resources continues to rise. To meet these requirements and support the project’s growth, the team relies on Qarnot, gradually expanding its usage to handle this increasing demand. Today, all their simulations run on Qarnot, and the partnership continues to strengthen.
"We began with test runs, validated performance, and transitioned all simulation workloads to Qarnot." Freek Bos · IT Manager
Gauss relies on Qarnot as its Cloud HPC provider for running all its simulations. Facing heavy workloads, ever-larger models, and increasing demand, the team explains: “As our simulation workloads grew in scale and complexity, we realized that a dedicated HPC partner would better support our performance and scalability goals, with the added benefit of closer collaboration compared to a general-purpose cloud platform.” With Qarnot, the company now has a stable and high-performing infrastructure, allowing them to run more tasks in parallel, increase computing throughput, and reduce time lost on technical issues: “Shorter turnaround times, stable and predictable performance.”
Beyond technical performance, Qarnot stands out for its responsive and attentive customer support. Freek Bos, IT Manager, emphasizes: “A partner who responds quickly and understands our needs,” along with the tangible benefits for their teams: “Better use of engineering time.” Cost and credit visibility, precise task tracking, and the ability to rapidly integrate new codes and tools allow the company to focus on innovation. This combination of HPC performance and reliable support makes Qarnot a trusted partner, capable of supporting Gauss’s growth and ambitions.
While Gauss focuses on fusion, Qarnot reduces the carbon footprint of its simulations by 80%. Furthermore, Qarnot controls the entire HPC cloud value chain -from designing low-carbon infrastructure to delivering high-performance and secure services- ensuring digital sovereignty in Europe, a critical factor for strategic sectors such as energy.
“The match in philosophy matters.” Freek Bos · IT Manager
Gauss Fusion has a dedicated team of engineers focused on advanced physics and fusion research. Their complex simulation workloads, including specialized codes like STELLOPT, Simsopt, FIELDLINES, BEAMS3D, and THRIFT, as well as industry-standard commercial software like ANSYS Fluent and ANSYS Mechanical, run effectively on Qarnot's infrastructure.
Standardized Environment through Containerization
The integration relies on containerization to ensure environmental consistency and simplify deployment. The process starts with a base docker image developed by Qarnot. It contains the packages required forMPI clustering using ROCE on Qarnot (RDMA over Converged Ethernet), which is a foundational software layer developed and provided by Qarnot. The ROCE base image is optimized for scalability across multiple compute nodes. On top of that base image, Gauss Fusion builds its own application layer (containing its physics codes and dependencies). This combination ensures that Gauss Fusion’s specific codes scale effectively on Qarnot’s infrastructure, removing the compute power bottleneck.
This custom image is then pushed by the Gauss Fusion administrator into Qarnot's shared container registry. This approach ensures that all team members pull and execute their jobs using the exact same verified environment, eliminating configuration drift and simplifying maintenance. This way, Gauss Fusion maintains complete ownership over its research codes, and can easily update, maintain, and fix its software tools.
For industry standard softwares like ANSYS Fluent and ANSYS Mechanical (see our catalog here), the integration is handled fully on Qarnot’s side. These solutions are treated as "on the shelf" tools, meaning the engineers benefit from a ready-to-use deployment for these applications, with a “Bring-Your-Own-Licence” (BYOL) model, where Qarnot provides a dedicated and secured connection to the client’s licence servers.
Tailored Account and Access Management
Balancing user autonomy with administrative oversight is key when optimizing teamwork. Here is how Gauss Fusion optimizes collaboration and controls resource usage. Each engineer has their own Qarnot account for submitting jobs via the Qarnot python SDK. This way, they can manage their computational tasks independently. Read-only access to other user accounts helps tracking project status across the team. This crucial feature also helps monitoring job progress, tracking resource utilization, and facilitating team collaboration without directly interfering with the execution environment. On the financial aspect, a central admin account manages the credits used to launch workloads on Qarnot with invoice delegation, meaning the budget is centralized and controlled for the whole team. Combined with Qarnot’s pay-as-you-go pricing model, this setup guarantees complete control, flexibility and predictability on the HPC billing, enabling Gauss Fusion to foresee and adapt budgets in advance.
