Sponsored Keynote: The Cloud (And Your $$$) is Not Infinite: Dynamic Scheduling at Every Layer
Talkat KubeCon + CloudNativeCon Europe 2025 · on Wednesday, 2025-04-02· at 09:27· Level 0 | ICC Auditorium·
summary
Kubernetes moves from static pools to on‑demand placement that claims to cut cost and handle scarce hardware by letting pods express hardware needs and preferences using dynamic resource allocation so workloads can get fractional GPU sharing or exclusive GPUs. Cluster autoscalers and tools like Carpenter and GK custom compute classes mix spot/reserved/ondemand priorities, nodes span regions when needed, and multicluster approaches such as multicluster orchestrator and MultiQ place AI training or inference across fleets for better availability and lower price.
AI-written summary · about
related talks
05mSponsored Keynote: Towards Clouds of AI Clusters
07mKeynote: Exploring the Frontiers of Cloud Native
05mSponsored Keynote: Kubernetes Everywhere
40mHow an AI driven approach reduces cloud cost and makes your Kubernetes infrastructure autonomous
31mAccelerate AI adoption: automated cloud efficiency
09mLightning Talk: Optimize the Ray Schedule and Autoscaling on Kubernetes Cloud: A Hetero...
10mKeynote: Enabling AI with Cloud Native
19mKeynote: Building Bridges: Cloud Native and High Performance Computing