The HPC Ready Architecture for AI and Data Analytics is a single environment that is tailored for a wide variety of HPC, AI, and high-performance data analytics workloads. The architecture provides a single infrastructure pool. You can assign the pool either to HPC jobs that are managed by an HPC resource manager like Slurm or to containerized AI and data analytics workloads that are orchestrated by the open-source Kubernetes container orchestration system.
The HPC Ready Architecture for AI and Data Analytics is supported by validation, interoperability testing, and benchmarking. The solution design uses a building-block approach, giving you flexibility in choosing the compute, network, storage, and software features that are best suited for the use cases and workloads that they will run in your environment. Customer conversations indicate that a no-one-size-fits-all approach works.
The following figures show the components and building blocks of the HPC Ready Architecture Solution Stack:
Figure 1. HPC Ready Architecture for AI and Data Analytics high-level architecture
Figure 2. HPC Ready Architecture for AI and Data Analytics building blocks