Benchmarking Your Distributed ML Training on the K8s Platform

This lightning talk explores how to benchmark distributed machine learning training on Kubernetes platforms. Discover the challenges of running ML training workloads on Kubernetes, including dynamic resource scaling, GPU scheduling, and efficient inter-node communication. Learn about recent advancements like KubeRay, Kubeflow, and Slurm integration that have expanded Kubernetes' capabilities for handling complex, large-scale ML training tasks. Explore the design and implementation of a benchmarking platform that provides actionable insights to improve throughput, scalability, and efficiency of distributed ML training workloads on Kubernetes.