Skip to content

Adoption & Ecosystem

Who uses it

The organisations below are listed as production users in the Volcano community adopters file. Use cases are not individually detailed there beyond the production designation.

OrganisationUse caseSource
Huawei CloudProduction batch schedulingadopters.md
TencentProduction batch schedulingadopters.md
BaiduProduction batch schedulingadopters.md
IQIYIProduction batch schedulingadopters.md
XiaohongshuProduction batch schedulingadopters.md
DiDiProduction batch schedulingadopters.md
iFlytekProduction batch schedulingadopters.md
Kingsoft CloudProduction batch schedulingadopters.md
ZoomProduction batch schedulingadopters.md

The list also names Bosszhipin, Ruitian Capital, and Momenta among production users.

Adoption signals

Measured from the GitHub API on 2026-06-25:

  • Stars: 5,699; forks: 1,415; open issues: 657.
  • Contributors: the GitHub contributors API runs to roughly 447 (anonymous included).
  • The CNCF incubation blog reports growth from 70+ to 350+ contributors and 5 to 50+ organisations between Sandbox acceptance and Incubation.

Ecosystem

Volcano integrates with batch and AI frameworks including Spark, Flink, Ray, PyTorch, TensorFlow, MindSpore, PaddlePaddle, Kubeflow, MPI/Horovod, Argo, and KubeGene, per the README. Apache Spark adopted Volcano as a built-in batch scheduler on Kubernetes (CNCF Spark blog). It supports heterogeneous devices such as GPU and NPU. For multi-cluster scheduling there is a separate volcano-sh/volcano-global federation scheduler.

Alternatives

AlternativeDiffers by
kube-scheduler (default)Schedules one pod at a time with no gang, queue, or fair-share. Volcano replaces or complements it for batch workloads.
kube-batchThe predecessor that Volcano's scheduler is built on; now largely inactive, with Volcano as its successor.
Apache YuniKorn (CNCF)Also a batch/queue scheduler for Kubernetes, centred on hierarchical queues. Volcano differentiates with the VolcanoJob CRD lifecycle and a wider plugin set (NUMA, device, topology).
Kueue (Kubernetes SIG)Focuses on job queueing and quota and defers placement to a scheduler. Volcano reaches down into the scheduling algorithm itself.
YARN and traditional HPC schedulersNot Kubernetes-native; require a separate resource manager.

Pick Volcano when the workload is gang-shaped and Kubernetes-native; pick Kueue when you only need queueing on top of an existing scheduler.