Cracking the code: AI workloads at scale on AKS
summary
Spot Ocean for AKS automates GPU and container capacity on Azure using AI-driven spot VM selection and application‑driven autoscaling, claiming up to 90% infrastructure cost reduction and a 45% cut in DevOps management time. The platform reports ~95% bin‑packing utilization and ~99.9% production coverage on Spot‑managed notebooks while preserving availability via SLA, and adds features like automatic right‑sizing, cluster roll/schedule, network cost analysis, and unified observability.
AI-written summary · about
chapters (6)
- 0:00AI workloads on AKS
- 2:50Challenges of AI workloads
- 5:20Day 2 CloudOps for AI workloads
- 6:27Spot Ocean for AKS use case
- 12:35Future of Spot platform on AKS
- 14:23Summary and next steps
related talks
31mAccelerate AI adoption: automated cloud efficiency
40mHow an AI driven approach reduces cloud cost and makes your Kubernetes infrastructure autonomous
20mHow are You Utilizing GPUs? Best Practices on Managing GPUs in Azure
01h 02mAKS Survival Pack: what to consider before going all-in with AKS
18mScaling Kubernetes-Native Systems with Event-Driven Patterns on Azure
29mAzure Container Instances(ACI) use-cases and roadmap
14mAnyscale on Azure: Build, Train, and Serve AI in Your Own Tenant
01h 01mDevOps with Azure Kubernetes Service