Simplifying AI Cluster Management with NVIDIA Base Command

Key Takeaways

Nvidia base command is the operating system of the accelerated data center providing everything you need to manage AI infrastructure backed by Nvidia in this video we'll show you how easy it is to manage a mix of Nvidia dgx systems and other nodes using the cluster management features of Base command base command provides a single pane of glass view across your entire cluster allowing you to drill down into any node and find out all pertinent information here we are inspecting a dgx node and can see information about the system including details about the operating system manufacturer system name and bios details as well as CPU and memory utilization health checks networking information some basic job overview details and the status of the gpus base command allows you to do initial provisioning and updating of nodes using the concept of a software image this image is the software blueprint and includes everything that will be installed on the nodes in your cluster the dgx operating system is provided as a default image and there are generic images for other Hardware the software images can be modified to meet your specific needs when it's time to distribute the image to the nodes it's simple to select a group of nodes and update the image across all of them with just a few clicks once the systems are provisioned you can check the health of the cluster and all the nodes it contains you can see a list of nodes in the cluster here we can go into the head node of the cluster and select latest health checks then we can see that there are a couple of items that need our attention administrators can choose to receive alerts when a health check fails via slack or email this functionality is very flexible allowing administrators to Define custom metrics even for data sources that aren't included by default allowing them to always be on top of the current status of the cluster administrators can also monitor the status of jobs within base command supporting popular workload Management systems such as kubernetes and slurm here we're creating a custom dashboard for a slurm cluster that lets us plot any of hundreds of metrics and get a real-time view of the data available metrics are shown in the panel on the left and we can simply drag and drop them onto the widgets here we will plot jobs complete versus jobs failed in our slurm partition we can add as many widgets as we want to visualize these metrics this is just a small sampling of how Nvidia base command can help you tap the full potential of your AI infrastructure click on the link in the description to learn more