Vivo's AI Computing Platform on Kubernetes

on YouTube ↗25m

" summary

Describes Vivo’s AI computing platform built on Kubernetes using kube-batch, Kubeflow and other open-source tools to schedule and run large-scale distributed model training and serving. Covers operating multiple production clusters with thousands of GPU nodes, how they improve GPU utilization and resource scheduling, the problems encountered, production best practices, and their contributions back to open-source projects.

AI-written summary · about

" chapters (5)

1. 0:00Start
2. 4:03Technology Selection
3. 7:28批调度 Kube-batch
4. 16:19Elastic Scaling with KEDA
5. 20:31Hybrid Cloud ACK

" related talks

• 17mDevOps for Private AI: A Scalable Platform Using KubeflowMarco Cristofanilli · Incontro DevOps Italia 2025
• 25mExtending Kubernetes for AI | Lessons Learned From Platform EngineeringAndrea Giardini, Lucy Sweet, Tim Wickberg, Etienne Dilocker, Susan Wu · KubeCon + CloudNativeCon Europe 2025
• 18mKeynote: Optimizing AI Workload Scheduling: Bilibili's Journey To an Efficient Cloud Native AI PlatformKevin Wang, Long Xu · KubeCon + CloudNativeCon China 2025
• 54mEconomics and Best Practices of Running AI/ML Workloads on KubernetesMaulin Patel · KubeCon + CloudNativeCon Europe 2019
• 43mRun Your AI Workloads and Microservices on Kubernetes More Easily and Efficiently...William Wang · KubeCon + CloudNativeCon + Open Source Summit China 2023
• 11mGPU‑Accelerated Workloads on KubeVirt: Scaling ML/AI in Kuberne...Amandeep Singh, Shivani Tiwari · KubeVirt Summit 2025
• 36mProduction GPU Cluster with K8s for AI and DL WorkloadsMadhukar Korupolu · KubeCon + CloudNativeCon Europe 2019
• 27mHarnessing Kubernetes for Scalable AI/ML WorkloadsPraneel Madabushini · Conf42 Machine Learning 2025