volcano-sh/volcano

A Cloud Native Batch System (Project under CNCF)

What it solves

Volcano is designed to address the limitations of the standard Kubernetes scheduler when handling high-performance batch and elastic workloads. It provides specialized scheduling capabilities required for AI, machine learning, deep learning, and Big Data applications that need efficient resource allocation and optimization across large-scale clusters.

How it works

Volcano extends and enhances the standard kube-scheduler to create a Kubernetes-native batch scheduling system. It integrates with a wide array of AI and HPC frameworks (such as PyTorch, TensorFlow, Ray, and Spark) to manage how these workloads are distributed and distributed across the cluster, incorporating best practices for high-performance computing (HPC) into a cloud-native environment.

Who it’s for

It is intended for engineers and organizations running large-scale AI/ML training, bioinformatics, genomics, and Big Data analytics platforms on Kubernetes, particularly those using frameworks like Kubeflow, Spark, or Flink.

Highlights

  • Broad Framework Integration: Robust support for PyTorch, TensorFlow, Ray, Spark, Flink, and MPI.
  • CNCF Incubating Project: Part of the Cloud Native Computing Foundation ecosystem.
  • High-Performance Focus: Specifically optimized for AI, deep learning, and LLM training workflows.
  • C-native Architecture: Built as a Kubernetes-native system to ensure seamless integration with the cluster's resource management.

Related

  • Project
    raids-lab/crater

    A Kubernetes-native control plane for shared AI computing clusters that simplifies GPU resource governance, workload scheduling, and asset management.

  • Project
    kai-scheduler/KAI-Scheduler

    A scalable Kubernetes scheduler that optimizes GPU resource allocation and fairness for large-scale AI and machine learning workloads.

  • Project
    kubeflow/trainer

    A Kubernetes-native distributed AI platform for scalable LLM fine-tuning and training across various frameworks like PyTorch, JAX, and XGBoost.

  • Project
    kubernetes-sigs/jobset

    A Kubernetes-native API for managing groups of Jobs as a single unit, specifically designed to simplify the deployment of distributed AI/ML training and HPC workloads.

  • Project
    fluid-cloudnative/fluid

    A Kubernetes-native distributed dataset orchestrator and accelerator that optimizes data access for AI and big data applications by providing dataset abstraction and elastic caching.