Uploaded August 2026 | Updated September 2026, 3 weeks ago
Don't miss out! Join us at our next KubeCon + CloudNativeCon events in Shanghai, China (8-9 September, 2026) and Salt Lake City, United States (Nov 9–12, 2026). Connect with our current graduated, incubating, and sandbox projects as the community gathers to further the education and advancement of cloud native computing. Learn more at kubecon.io
Beyond Single-Cluster Limits: Scaling GPU Workloads Across Kubernetes With Virtual Nodes - Kunal Das, Cast AI & Esmira Bayramova, Kimchi
Our ML platform team hit a wall when GPU demand became unpredictable. Fine-tuning jobs queued for hours on busy days while expensive on-prem GPUs sat idle on others. Adding cloud GPU clusters solved capacity but fractured our workflow , developers juggled multiple kubeconfigs, Kueue couldn't see cross-cluster queues,workloads couldn't failover because "multi-cluster" was really just isolated clusters with shared dashboards.
In this talk, we walk through how we used Liqo's virtual node pattern to unify 5 heterogeneous clusters , mixing on-prem A100/ H100 nodes with cloud spot GPU instances from cloud providers , into a single schedulable topology. We'll share how the vanilla Kubernetes scheduler handles placement using standard node selectors and affinities, how Cilium's cluster mesh compared to Liqo's WireGuard-based network fabric for cross-cluster pod traffic, and how we integrated Kueue for unified job queuing across the virtual cluster.
Don't miss out! Join us at our next KubeCon + CloudNativeCon events in Shanghai, China (8-9 September, 2026) and Salt Lake City, United States (Nov 9–12, 2026). Connect with our current graduated, incubating, and sandbox projects as the community gathers to further the education and advancement of cloud native computing. Learn more at kubecon.io
Beyond Single-Cluster Limits: Scaling GPU Workloads Across Kubernetes With Virtual Nodes - Kunal Das, Cast AI & Esmira Bayramova, Kimchi
Our ML platform team hit a wall when GPU demand became unpredictable. Fine-tuning jobs queued for hours on busy days while expensive on-prem GPUs sat idle on others. Adding cloud GPU clusters solved capacity but fractured our workflow , developers juggled multiple kubeconfigs, Kueue couldn't see cross-cluster queues,workloads couldn't failover because "multi-cluster" was really just isolated clusters with shared dashboards.
In this talk, we walk through how we used Liqo's virtual node pattern to unify 5 heterogeneous clusters , mixing on-prem A100/ H100 nodes with cloud spot GPU instances from cloud providers , into a single schedulable topology. We'll share how the vanilla Kubernetes scheduler handles placement using standard node selectors and affinities, how Cilium's cluster mesh compared to Liqo's WireGuard-based network fabric for cross-cluster pod traffic, and how we integrated Kueue for unified job queuing across the virtual cluster.
