How Volcano Is Shaping Cloud‑Native Batch Computing in CNCF

On April 7, the Cloud Native Computing Foundation (CNCF) announced that Volcano, the industry’s first cloud‑native batch computing project contributed by Huawei Cloud, has been promoted to a CNCF incubating project.

Huawei Cloud CTO Zhang Yuxin said that Volcano integrates Huawei Cloud’s experience in cloud‑native, AI, big data and HPC, helping enterprises efficiently merge cloud‑native technology with data assets, unlock data dividends, and accelerate digital and intelligent transformation.

CNCF CTO Chris Aniszczyk noted that batch processing simplifies complex Kubernetes workloads, especially in AI and big‑data domains, and that Volcano makes Kubernetes a world‑class tool for scientific research and high‑performance computing.

Volcano was open‑sourced in June 2019, entered the CNCF sandbox in April 2020, and has since been applied in AI, big data, and genomics scenarios by companies such as Tencent, iQIYI, Xiaohongshu, Mogujie, Vipshop, Pengcheng Laboratory and Ruian Investment.

Since joining CNCF, the Volcano community has attracted 26,000 global developers, earned 2.3k stars and over 530 forks, and received contributions from Google, Facebook, Amazon, Red Hat, Huawei, Baidu, Tencent and others. It has won awards such as “China Excellent Open‑Source Project” and the 2021 OSCAR Peak Open‑Source Community and Project Award, and its job‑management capabilities have been incorporated into the “High‑Performance Computing Cloud Platform Standard” drafted by the China Academy of Information and Communications Technology.

Industry users report that Volcano simplifies and accelerates large‑scale AI and data workloads in cloud‑native environments, reducing resource costs and improving job performance.

Key roadmap items for 2022 include:

Cross‑cloud, cross‑cluster scheduling : building a hierarchical scheduler with global resource view and policies such as cost‑first or efficiency‑first.

Mixed‑mode offline jobs : improving resource utilization through business‑aware scheduling, oversubscription, isolation and preemption.

Elastic scheduling : enhancing min/max based scheduling, job awareness and resource preemption for elastic training and spot instances.

GPU virtualization : enabling multi‑container GPU sharing and improving compute and memory isolation.

Fine‑grained resource management : extending queue mechanisms (hierarchical queue, policy per queue, plugin per queue) for more precise sharing.

Workflow management : developing lightweight, high‑performance orchestration via the JobFlow sub‑project.

Dynamic scheduling based on real workloads : using monitoring data to rebalance and reschedule jobs.

Huawei Cloud has been a pioneer in cloud‑native technologies since 2015, contributing projects such as KubeEdge, Karmada and Istio, and continues to promote cloud‑native adoption through white papers and a global cloud‑native exchange platform.