Key Benefits

Utilizing Containers for High Performance Computing (HPC) is key for transforming traditional static AI training and CAE simulation workflows into a dynamic demand driven environment. This main differentiator from current setups will be highlighted and favor for a suitable architecture will be shown during the assessment. This approach allows boosting your business far beyond the current state and leveraging efficient distributed HPC Grids thus empowering the infrastructure and save costs.

Challenges

Although Docker Containers and Amazon EKS are well established technologies adaption for High Performance workflows is challenging:

• Parallel MPI jobs consist of multiple concurrent Containers running on multiple hosts that need to be able to communicate with minimum latency with each other
• Solver codes need to access low-latency networks --- from within the Container
• HPC-enabled Container runtimes like Podman or Singularity have various advantages and disadvantages that need to be known before selecting a platform
• Container orchestration has to be integrated with the HPC batch system (Slurm, etc.)

Key Activities

Our Container architects and HPC experts will work with your specialist teams and provide

• Analysis of the workflow processes to choose proven best practice architectures for HPC Containers
• Insight into the performance impact of the Container layer technology
• Guidelines to choose the best suited Container runtime platform (Docker, Podman, Singularity, etc.)
• Recommendations on focused Proof of Concepts to make sure a Container migration integrates with existing data management and engineering workflows

Why to choose Atos science + computing

Atos science + computing bespoke IT services for HPC assure these environments are fully integrated with existing customers’ workflows, specialized application and data management solutions. The offering is based on the bespoke expertise of designing and running streamlined HPC, CAE und Data Analytics IT environments. Decades of successful joined customer projects in the industry are the foundation to assure an optimized use of AWS dedicated HPC resources. Being a trusted advisor for HPC Atos science + computing assures cloud adoption decisions based on hands-on expertise and viable business cases.