Project Lightning Talk: Next-Gen AI Orchestration With Volcano On Kubernetes - Zhonghu Xu

This talk features Zhonghu Xu, a Principal Engineer, discussing the advanced AI orchestration capabilities of Volcano on Kubernetes. The speaker explains how Volcano serves as a Unified Scheduling Platform, addressing the challenges of fragmented AI infrastructure by facilitating efficient training, inference, and agent workloads. Key topics include the evolution of Volcano beyond batch processing to coordinate various schedulers and workload types, support for enterprise large language model serving with Kthena and vLLM, and enhancements to agent runtime using AgentCube. The session also highlights infrastructure modernization strategies involving DRA, HyperNode discovery, GPU sharing, and heterogeneous pooling.