Sponsored Keynote: The Weight of Data: Rethinking Cloud-Native Systems for the Age of AI
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 09:32· Level 0 | ICC Auditorium·
summary
Argues cloud-native must shift from stateless to state-aware systems because AI agents produce high-volume, shared, mutable context and traditional failure modes (crash/restart) are insufficient when agents return wrong outputs. Calls out Kubernetes primitives—PersistentVolumes, StatefulSets, DaemonSets—and projects VEST, Rook, K native and streamy as foundations but insufficient at AI scale. Recommends NUMA-aware scheduling, GPU and topology-aware placement, an LLM gateway for routing heavy inference, plus OpenTelemetry-based correlation and a new fault-tolerance model for distributed, high-throughput state.
AI-written summary · about
related talks
32mWhy we need to think about AI and Cloud Native NOW
05mSponsored Keynote: Towards Clouds of AI Clusters
23mThe State of Cloud Native: The Shift Towards AI
16m[Opening Keynote] Making Distributed Computing Easy
20mCloud-Native AI: Architectural Patterns for Enterprise Success
27mCloud Native Artificial Intelligence - Top to Bottom
16mKeynote: An Open-Source Stack for AI Compute: Kubernetes + Ray + PyTorch + vLLM
04mClosing Remarks