Recap of Kubernetes Community Day 2024 Jakarta: Generative AI, eRDMA, Container Security, and Observability

On November 30, 2024, the Kubernetes Community Day (KCD) took place in Jakarta, Indonesia, bringing together over 350 developers and technologists for keynote talks and hands‑on sessions. Alibaba Cloud, as a sponsor, contributed four container and observability experts who delivered one main forum presentation and three technical sub‑forum talks.

In the main forum, container specialist Xu Zhihao demonstrated how to run generative AI applications on Kubernetes using Alibaba Cloud Container Service for Kubernetes (ACK) combined with open‑source projects such as Kubeflow/Arena for workload management, KServe for inference orchestration, MLflow for model metadata, and Fluid for rapid data access, addressing scalability, model management, user experience, and data latency challenges.

During the first sub‑forum, Li Bokang introduced Alibaba Cloud's Elastic Remote Direct Memory Access (eRDMA) network, highlighting its high‑throughput, low‑latency characteristics, seconds‑level large‑scale RDMA group formation, and the open‑source eRDMA Controller component designed for ACK to simplify pod‑level network configuration and boost performance.

The second sub‑forum, presented by Wu Hengyu, focused on container security. Leveraging ACK’s security capabilities, Alibaba Cloud built an OCI artifact signing verification solution that integrates Gatekeeper policy enforcement, the Ratify component, and the Notation‑AlibabaCloud‑Secret‑Manager plugin (linked with Alibaba Cloud KMS) to ensure image integrity, enable dynamic signature verification, and provide real‑time admission control.

In the final sub‑forum, Cao Jian shared Alibaba Cloud’s observability stack based on OpenTelemetry. He explained how the unified data collection, processing, and storage approach combines OpenTelemetry standards with tools like Jaeger and Prometheus, and incorporates AI‑driven anomaly detection, root‑cause analysis, automated remediation, and Qwen‑powered semantic search to achieve comprehensive cloud‑native monitoring.