Keynote: Optimizing AI Workload Scheduling: Bilibili's Journey To an Efficient Cloud Native AI Platform

on YouTube ↗18m

" summary

Describes Bilibili's approach to scheduling diverse AI workloads (training, inference, video processing) across multiple Kubernetes clusters while meeting SLAs, low startup latency, and maintaining QoS. Explains how they adapted CNCF projects like Karmada and Volcano, integrated KubeRay for online/offline AI tasks, and used hybrid online/offline and tidal scheduling techniques to boost resource efficiency.

AI-written summary · about

" related talks

• 10mKeynote: Enabling AI with Cloud NativeJorge Castro · AI.dev: Open Source GenAI & ML Summit North America 2023
• 10mKeynote: Enabling AI with Cloud NativeJorge Castro · Cassandra Summit 2023
• 06mSponsored Keynote: Building Your AI Factory: Wh...Andreea Munteanu · Cloud Native AI + Kubeflow Day 2026
• 15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?Saiyam Pathak · KubeCon + CloudNativeCon India 2026
• 43mRun Your AI Workloads and Microservices on Kubernetes More Easily and Efficiently...William Wang · KubeCon + CloudNativeCon + Open Source Summit China 2023
• 30mVolcano: A Unified Scheduling Platform for Cloud Native AIChen Zicong, DongYang Wang · KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026
• 14mKeynote: Building the Infrastructure that Powers the Future of AIVicki Cheung, Jonas Schneider · CloudNativeCon + KubeCon Europe 2017
• 06mKeynote: The Cloud Native Ecosystem in ChinaTao Ma · KubeCon + CloudNativeCon China 2018