CHALLENGE

Optimizing infrastructure for R&D efficiency

In a research and development initiative requiring high computational power, Prometeon needed to run iterative simulations and complex analyses with greater agility and efficiency. To further enhance the R&D capabilities, they aimed to evolve the infrastructure to efficiently scale complex simulations, enabling faster responses to peak workloads and accelerating the deployment of new configurations and software updates.

Enhancing computational agility with BIP xTech and Google Cloud

To address these challenges, BIP xTech supported Prometeon with continuous delivery pipelines, enabling them to test and refine the optimal setup. This process involved experimenting with different machine types, CPU architectures, machine sizes, and features like multithreading, network drivers, and more.

Additionally, Google Cloud provided assistance during the Slurm tuning phase—the orchestrator responsible for spinning up, scheduling, and tearing down VMs.

Following an initial adjustment period, the benchmarking results demonstrated strong performance, leading Prometeon to fully adopt the configuration thanks to significantly accelerated computations.

OUR SOLUTION

Scalable and resilient Cluster Toolkit

BIP xTech assisted Prometeon in implementing a high-performance computing (HPC) solution. We leveraged the Cluster Toolkit on Google Cloud Platform (GCP), designed specifically to improve R&D efficiency. This toolkit allows users to define the architecture of an HPC cluster, including controllers and auxiliary resources, via a simple YAML configuration file. The application generates Infrastructure as Code (IaC) files, primarily in HCL (Terraform), streamlining infrastructure deployment.

Streamlined infrastructure management with Cluster Toolkit

Cluster Toolkit also simplifies the creation of custom images using Packer, which are natively stored in the Compute Engine image service. It adopts a “cattle” service model for infrastructure management, where resources are treated as replaceable. Under this model, failed resources are terminated and replaced automatically. Updates, such as new machine types or base images, involve deleting and recreating the cluster with the latest features seamlessly.
This approach enhances scalability, resilience, and efficiency, making it ideal for Prometeon’s dynamic R&D needs.

RESULTS

Driving R&D efficiency

The implementation of our solution led to remarkable outcomes for Prometeon’s R&D operations:

• Infrastructure costs were reduced by 30%.
• Pure computation time decreased by 35%.
• A follow-up discussion nearly a year after implementation revealed a significant productivity boost across the R&D team, largely due to the user-friendly, highly customizable environment.

These improvements allowed Prometeon to optimize workflows, enabling their team to focus on innovation rather than infrastructure management.

Conclusion

The Cluster Toolkit on Google Cloud Platform empowered Prometeon to overcome their scalability and efficiency challenges. By leveraging cloud-native features like instance templates and startup scripts, the solution provided a secure, scalable, and highly manageable environment. This not only simplified collaboration among regional teams but also unlocked new levels of productivity and cost savings for their R&D efforts.