Canonical and NVIDIA have jointly developed this reference architecture guide for the infrastructure required for machine learning (ML) use cases. It focuses specifically on opensource ML software, including MicroK8s and Charmed Kubeflow, both delivered by Canonical and running on the NVIDIA EGX platform, which includes NVIDIA Certified Systems provided by various hardware vendors.
Canonical MicroK8s is a Kubernetes distribution certified by the Cloud Native Computing Foundation (CNCF). Ongoing collaboration between NVIDIA and Canonical ensures continuous validation of test suites, enabling data scientists to benefit from infrastructure designed for AI at scale using their preferred MLOps tooling, such as Charmed Kubeflow.
The guide covers hardware specifications, tools, services, and provides a step-by-step guide for setting up the hardware and software required to run ML workloads. It also delves into other tools used for cluster monitoring and management, explaining how all these components work together in the system. At the end of it, users will have a stack that is able to run AI at the edge.