After clicking “Watch Now” you will be prompted to login or join.
Training and Inferencing at Scale, Across Node and Cluster Borders with Optimized Software and Hardware Stack
Zvonko Kaiser, Red Hat
GTC 2020
The demand for computational power for AI/ML workloads keeps rising. While it is easy to burst out work into the cloud, costs can quickly add up for every spun-up instance. Learn how you can reduce and efficiently manage these computational costs by optimizing the hardware and software stack to fully leverage the features a node in a cluster provides. We'll discuss the latest Kubernetes features for Pod hardware affinity and NUMA awareness, as well as leveraging operators for your AI/ML deployments.