Reduce Time to Market and Capital Spend using Software Operators
summary
Claims a concrete cut in time-to-market and capital spend by packaging Slurm as an operatorized distribution and automating day-two ops for HPC. Uses Omnivector Slurm Distribution built as Juju charms with the Canonical Operator Framework, deploys to LXD, AWS, GCP, Azure and VMware, submits jobs via Slurm REST API and validates output with PowerView. Replaces heavy Juju relation chatter with an NCD key-value store, supports clusters measured in hundreds of nodes, and reports ops as ~60% of HPC overhead.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:28James Beatty
- 0:52About OmniVector
- 1:48What makes a computer high performance
- 2:54Where we are today
- 4:59Reducing Time to Market
- 5:18Overspending
- 5:40Operations
- 7:00Cloud
- 8:16OSD
- 8:39Linux Utility Resource Manager
- 9:33Use Case for Software Operators
- 10:13Cloud Coverage
- 10:34Models Operating
- 12:07Demo
- 13:31Job
- 14:09Power Review
- 15:34NCD
- 18:02Use Cases
- 20:38Custom Integrations
related talks
18mThe Operations Dividend
39mManaged CF-as-a-Service: An Operator's Cookbook
03mSponsored Keynote: Redefining MLOps at Scale on Top of Kubernetes
05mTaming the Metal: Platform Engineering for HPC
45mIT Revolution: SLOs, DevOps, & Software-Defined Datacenter - Cloud Efficiency
48mCombating the struggles of modern cloud-native java developers
30mBeating the competition through faster time to market
27mContaining the Chaos While Embracing Kubernetes Based Technology in Finance