Project Lightning Talk: Scheduling AI Workload Among Multiple Clusters
Lightningat KubeCon + CloudNativeCon Europe 2025 · on Tuesday, 2025-04-01· at 11:52· Platinum Suite | Level 3·
summary
Open Cluster Management uses a hub‑and‑spoke inventory, a placement CRD and add‑ons to map cluster capabilities; a GPU/TPU resource collector addon computes an AddonPlacementScore so placement filters pick clusters with the most accelerators. The demo wires that score into Q’s multiQ config so batch AI jobs land on clusters/nodes with available GPUs, and the same placement approach supports federated‑learning residency requirements across fleets. Implementations are shown working end‑to‑end with cluster registration, scoring, and job placement.
AI-written summary · about
related talks
05mScheduling at the Edge of Reason: Multi-Cluster AI & OCM
30mScore-Driven Multi-Cluster Management: An Evaluation Framework for…
22mBeyond Single-Cluster Limits: Scaling GPU Workloads Across Kubernetes With…
21mDesigning Scalable Networks for Large AI Clusters: Challenges and Key Insights
07mLightning Talk: Environmentally Sustainable AI via Power-Aware Batch Scheduling
21mMulti Cluster Magics With Argo CD and Cluster Inventory
24mCluster Management for Large Scale AI and GPUs: Challenges and Opportunities
37mCo-Location of CPU and GPU Workloads with High Resource Efficiency