Transcript
In this screencast, we will preview the upcoming Once Learned feature that brings HPC Cluster Management into the platform. Let's define the problem. Every time a researcher needs an HPC cluster, they have to deal with multiple layers and involve multiple teams. There are no common tools or workflows, and overall lack of visibility. OpenEmily is solving the challenge by introducing the Once Learned component. From provisioning the HPC clusters to monitoring and operations. Once Learned is a new section inside SunStorm that brings HPC Cluster Management into the platform. Slurm clusters, provisioned and monitored from the same place you manage your virtual machines. When you select a cluster, you get a live snapshot. Status, active partitions, GPU load, queued jobs, all at a glance. The Node tab shows each compute node's load in real time. Click any node for the full picture, GPU specs, partition, operating system, and optime. The Job tab gives you a full queue, running, queued, completed, with user, partition, duration, and a node count at a glance. Storage is modeled as part of the cluster. Home directories, scratch, and shared data, each with its mount path, server, and capacity. No guessing what's mounted where. Need more capacity? Hit scale, set the new node count, and confirm. Once Learned handles the rest. Provisioning a new cluster starts with a three-step wizard. Step one, cluster identity, Slurm version, and control plane. Step two, the node configuration. Pick your GPU instance type, define your partitions, set cores and memory per node, the building blocks of your HPC queue. Step three, storage. Mount paths and server targets for every file system the cluster needs. Set it once, once Learned takes care of the rest. And then just click the create button. One wizard, three steps, a fully configured Slurm cluster, managed inside OpenAbylla. And this concludes this feature preview demonstration. Thank you for watching and see you in the next screencast.
