Home > Servers > Modular Servers > White Papers > Reference Architecture: Machine Learning Containers on PowerEdge MX and VMware Cloud Foundation 4.0 with Tanzu > Overview
CRITICAL REQUIREMENTS FOR A MACHINE LEARNING INTEGRATED PLATFORM:
Kubeflow is a known leader in ML management and runtime platforms, running on top of Kubernetes. With an introduction of stable Kubeflow release (Kubeflow version 1.0) and release of production-ready Kubernetes clusters support in VMware vSphere 7.0 our goal was to make sure that we could use these two products together to deliver a Machine Learning/Data science platform to the end-users.
ML practitioners use diverse tools, and Kubeflow aims to cover different needs with a comprehensive set of instruments for prototyping, training and serving. Kubeflow comes as universal distribution. Thus, we have updated some components to support specific technologies like Intel DAAL version of XGBoost for serving models via Seldon deployments to create a machine learning deployment process with peak efficiency, minimal risk, and the shortest time-to-value.
Note: While Kubeflow 1.0 is not marked “compatible” officially with Kubernetes version 1.16 and higher – it did not show any issues during our operations.