From f1f5310c19bc54dcee387ef3f071cc15635499bc Mon Sep 17 00:00:00 2001 From: Aaradhy Chinche Date: Mon, 19 Jan 2026 16:28:57 +0530 Subject: [PATCH 1/2] docs: add Hierarchical Queue feature introduction Signed-off-by: Aaradhy Chinche --- .../docs/key-features/hierarchical-queue.md | 68 +++++++++++++++++++ 1 file changed, 68 insertions(+) create mode 100644 content/en/docs/key-features/hierarchical-queue.md diff --git a/content/en/docs/key-features/hierarchical-queue.md b/content/en/docs/key-features/hierarchical-queue.md new file mode 100644 index 00000000..40eb48b5 --- /dev/null +++ b/content/en/docs/key-features/hierarchical-queue.md @@ -0,0 +1,68 @@ ++++ +title = "Hierarchical Queue" +linktitle = "Hierarchical Queue" +type = "docs" +toc = true + +[menu.docs] + parent = "Key Features" + weight = 1 ++++ + +## Overview + +Hierarchical Queue is a core scheduling feature in Volcano that allows users to organize workloads into multi-level queues. It enables fine-grained resource management and fair sharing across teams, applications, or business units. + +By structuring queues hierarchically, cluster administrators can better control how resources are allocated and ensure that critical workloads receive appropriate priority. + +--- + +## Why Hierarchical Queue Is Needed + +In shared Kubernetes clusters, multiple teams and applications often compete for limited resources. Without proper queue management, some workloads may starve while others monopolize resources. + +Hierarchical queues address this problem by: +- Providing isolation between different teams or workloads +- Enforcing resource quotas at different levels +- Improving overall cluster fairness and utilization + +--- + +## Queue Models in Volcano + +Volcano supports different queue models to meet various scheduling requirements. + +### Capacity Queue + +A capacity queue defines the **maximum amount of resources** that a queue can use. If a queue reaches its capacity limit, additional workloads must wait until resources are freed. + +Capacity queues are useful when: +- Strong resource isolation is required +- Teams have fixed resource budgets +- Predictable resource usage is important + +--- + +### Proportion Queue + +A proportion queue allocates resources based on a **percentage of the total cluster capacity**. Resources are distributed proportionally among queues according to their configured shares. + +Proportion queues are useful when: +- Workloads are dynamic +- Fair sharing is preferred over strict limits +- Clusters need to adapt to changing demand + +--- + +## Typical Use Cases + +Hierarchical queues are commonly used in scenarios such as: +- Multi-tenant clusters shared by multiple teams +- AI and big data platforms with diverse workloads +- Enterprises requiring both fairness and isolation in scheduling + +--- + +## Next Steps + +This page provides a high-level introduction to hierarchical queues. Detailed configuration examples and advanced scheduling policies will be covered in follow-up documentation. From 0b7e34d344afd24c3f5f1c5749315bd6daf99eb4 Mon Sep 17 00:00:00 2001 From: Aaradhy Chinche Date: Mon, 19 Jan 2026 16:05:22 +0530 Subject: [PATCH 2/2] docs: add Key Features section to documentation Signed-off-by: Aaradhy Chinche --- content/en/docs/key-features/_index.md | 31 ++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) create mode 100644 content/en/docs/key-features/_index.md diff --git a/content/en/docs/key-features/_index.md b/content/en/docs/key-features/_index.md new file mode 100644 index 00000000..b2499feb --- /dev/null +++ b/content/en/docs/key-features/_index.md @@ -0,0 +1,31 @@ ++++ +title = "Key Features" +linktitle = "Key Features" +type = "docs" +toc = true + +[menu.docs] + parent = "scheduler" + weight = 2 ++++ + +## Overview + +Volcano is a cloud-native batch system built on Kubernetes, designed to run high-performance workloads such as AI, Big Data, and HPC jobs efficiently. This section introduces the key features that make Volcano powerful and flexible for diverse scheduling scenarios. + +## Core Capabilities + +### Unified Scheduling on Kubernetes +Volcano is fully compatible with Kubernetes and provides unified scheduling capabilities for batch workloads, online services, and mixed workloads within a single cluster. + +### Gang Scheduling +Volcano supports gang scheduling, ensuring that a group of related pods are scheduled together or not at all. This is critical for distributed workloads such as MPI and deep learning training. + +### Hierarchical Queues +Hierarchical queues allow users to organize workloads into multi-level queues with fine-grained resource control, improving fairness and utilization across teams and applications. + +### Colocation of Online and Offline Workloads +Volcano enables colocation of latency-sensitive online services and resource-intensive offline jobs, improving overall cluster utilization. + +### Extensible Scheduling Plugins +Volcano provides a rich plugin system that allows users to customize scheduling behavior for different scenarios, such as GPU scheduling, topology-aware scheduling, and resource fairness.