# kueue

Kubernetes-native job queueing system that manages quotas and job admission.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Gang scheduling and topology-aware placement for AI training in Kubernetes

DevFeed: [Gang scheduling and topology-aware placement for AI training in Kubernetes](<https://devfeed.tech/articles/gang-scheduling-and-topology-aware-placement-for-ai-training-in-kubernetes-130586.md>)

Original publisher: [Read original article](<https://www.cncf.io/blog/2026/11/12/gang-scheduling-and-topology-aware-placement-for-ai-training-in-kubernetes/>)

Author: David Lentz, DataDog

Published: 2026-11-12T12:00:00Z

Content type: article

Language: en

Sources: [Cloud Native Computing Foundation](<https://devfeed.tech/sources/cloud-native-computing-foundation.md>)

Topics: [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [NCCL](<https://devfeed.tech/topics/nccl.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [blog](<https://devfeed.tech/tags/blog.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [topology](<https://devfeed.tech/tags/topology.md>)

### AI overview

Distributed AI training needs both topology-aware placement and gang scheduling: Kueue assigns pods to a suitable labeled topology domain at admission, while the Coscheduling plugin waits until the required pod group can be scheduled together. The article recommends validating the pipeline across layers: check Kueue queue and eviction metrics and the Workload's topology assignment; inspect PodGroup state and events for gang assembly; then correlate GPU, PCIe, network, and training throughput signals. Kueue depends on topology labels being supplied by an administrator or provider, and healthy infrastructure metrics alone do not prove effective training progress. Kubernetes is adding native scheduling features, but the excerpt describes them as early and incomplete: gang scheduling and workload-aware preemption are beta in v1.37 but disabled by default, while native topology-aware scheduling remains alpha. Kueue and Coscheduling are presented as the current production path.

### Source excerpt

Distributed AI training workloads impose complex scheduling requirements that Kubernetes's built-in scheduler can't meet. Kubernetes schedules pods individually and independently, but distributed training introduces two requirements that break this model: Pods must land on hardware with...

## How Kueue schedules Spark jobs fairly on Red Hat OpenShift AI

DevFeed: [How Kueue schedules Spark jobs fairly on Red Hat OpenShift AI](<https://devfeed.tech/articles/how-kueue-schedules-spark-jobs-fairly-on-red-hat-openshift-ai-116313.md>)

Original publisher: [Read original article](<https://developers.redhat.com/articles/2026/11/04/how-kueue-schedules-spark-jobs-fairly-on-red-hat-openshift-ai>)

Author: Vedant Deshpande

Published: 2026-11-04T13:16:25Z

Content type: article

Language: en

Sources: [Red Hat](<https://devfeed.tech/sources/red-hat.md>), [Red Hat Developer](<https://devfeed.tech/sources/red-hat-developer.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [Apache Spark](<https://devfeed.tech/topics/spark.md>)

Tags: [admission](<https://devfeed.tech/tags/admission.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [jobs](<https://devfeed.tech/tags/jobs.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [openshift-ai](<https://devfeed.tech/tags/openshift-ai.md>), [red-hat-openshift](<https://devfeed.tech/tags/red-hat-openshift.md>), [schedules](<https://devfeed.tech/tags/schedules.md>), [spark](<https://devfeed.tech/tags/spark.md>), [spark-operator](<https://devfeed.tech/tags/spark-operator.md>)

### AI overview

Kueue controls when SparkApplications may start by suspending each submission, evaluating its declared CPU and memory against a ClusterQueue, and admitting it to the Spark Operator when quota is available. In the example, a self-managed queue reserves 3 CPUs and 4 GiB: one app with a driver and two executors fits, while a second waits without creating pods and starts after the first finishes. Setup requires the external Red Hat build of Kueue, SparkApplication framework integration and RBAC, a ResourceFlavor/ClusterQueue/LocalQueue, and a queue-name label on every app. The excerpt warns that preemption deletes driver and executor pods without checkpointing, dynamic executor allocation is unsupported, and quota reflects requested rather than actual resource use. It recommends fixed executor counts, explicit resource requests, and checking Workload conditions to distinguish a quota wait from a configuration problem.

### Source excerpt

Table of contents: How Kueue and the Spark Operator divide the work What you will build Set up the integration The post How Kueue schedules Spark jobs fairly on Red Hat OpenShift AI appeared first on Red Hat Developer.

## Volcano v1.16 Released: NamespaceQueue, fairshare, and a broader HyperNode ecosystem

DevFeed: [Volcano v1.16 Released: NamespaceQueue, fairshare, and a broader HyperNode ecosystem](<https://devfeed.tech/articles/volcano-v1-16-released-namespacequeue-fairshare-and-a-broader-hypernode-ecosystem-105924.md>)

Original publisher: [Read original article](<https://www.cncf.io/blog/2026/10/28/volcano-v1-16-released-namespacequeue-fairshare-and-a-broader-hypernode-ecosystem/>)

Author: Volcano Maintainers

Published: 2026-10-28T20:45:08Z

Content type: release

Language: en

Sources: [Cloud Native Computing Foundation](<https://devfeed.tech/sources/cloud-native-computing-foundation.md>)

Topics: [releases](<https://devfeed.tech/topics/releases.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [Cloud](<https://devfeed.tech/topics/cloud.md>)

Tags: [administrator](<https://devfeed.tech/tags/administrator.md>), [blog](<https://devfeed.tech/tags/blog.md>), [ecosystem](<https://devfeed.tech/tags/ecosystem.md>), [namespaces](<https://devfeed.tech/tags/namespaces.md>), [plugin](<https://devfeed.tech/tags/plugin.md>), [queue](<https://devfeed.tech/tags/queue.md>), [queues](<https://devfeed.tech/tags/queues.md>), [quotas](<https://devfeed.tech/tags/quotas.md>), [resource](<https://devfeed.tech/tags/resource.md>), [volcano](<https://devfeed.tech/tags/volcano.md>)

### AI overview

Volcano v1.16.0 adds namespace-scoped queue hierarchies and fairshare scheduling for tenants that share a queue. NamespaceQueue lets administrators authorize namespaces against cluster Queues while tenants manage quotas and child queues; it is disabled by default and initially requires hierarchical capacity. The fairshare plugin tracks decaying resource-seconds by namespace to influence pending-job order, optionally gate admission, and optionally persist state; it does not preempt running jobs. Other additions include a disabled-by-default Unschedulable Job Cache, which uses plugin event hints to skip repeated checks and retry on relevant changes, with a five-minute default watchdog; unsupported rejection types continue to be checked each session. PodGroup anti-affinity can separate groups across HyperNode domains during allocation, but offers no isolation guarantee for BestEffort backfill placements. A standalone HyperNode controller lets other schedulers reuse topology discovery and supernode management through the API. The release also adds Kubernetes 1.36 support and other scheduling and accelerator improvements.

### Source excerpt

Volcano v1.16.0 is available . v1.16.0 introduces NamespaceQueue and the fairshare plugin. Tenants can use hierarchical queues within their own namespaces to organize quotas and queue hierarchies within administrator-authorized resource pools. Fairshare brings historical resource consumption...

## How We Run 250,000 Automated Tests a Day on Kubernetes

DevFeed: [How We Run 250,000 Automated Tests a Day on Kubernetes](<https://devfeed.tech/articles/how-we-run-250-000-automated-tests-a-day-on-kubernetes-102335.md>)

Original publisher: [Read original article](<https://hackernoon.com/how-we-run-250000-automated-tests-a-day-on-kubernetes?source=rss>)

Author: lokeshsuri

Published: 2026-10-27T11:00:02Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [Test automation](<https://devfeed.tech/topics/test-automation.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [Amazon Elastic Kubernetes Service](<https://devfeed.tech/topics/amazon-elastic-kubernetes-service.md>), [Platform Engineering](<https://devfeed.tech/topics/platform-engineering.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [dynamic test splitting](<https://devfeed.tech/topics/dynamic-test-splitting.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [amazon-eks](<https://devfeed.tech/tags/amazon-eks.md>), [audit](<https://devfeed.tech/tags/audit.md>), [automated](<https://devfeed.tech/tags/automated.md>), [certification](<https://devfeed.tech/tags/certification.md>), [cicd](<https://devfeed.tech/tags/cicd.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [devops](<https://devfeed.tech/tags/devops.md>), [eks](<https://devfeed.tech/tags/eks.md>), [financial-services](<https://devfeed.tech/tags/financial-services.md>), [flaky-tests](<https://devfeed.tech/tags/flaky-tests.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [platform-engineering](<https://devfeed.tech/tags/platform-engineering.md>), [policy](<https://devfeed.tech/tags/policy.md>), [quotas](<https://devfeed.tech/tags/quotas.md>), [sharding](<https://devfeed.tech/tags/sharding.md>), [software-testing](<https://devfeed.tech/tags/software-testing.md>), [test-automation](<https://devfeed.tech/tags/test-automation.md>), [testing](<https://devfeed.tech/tags/testing.md>), [tests](<https://devfeed.tech/tags/tests.md>)

### AI overview

The article describes a centralized platform running about 250,000 tests daily across four Amazon EKS clusters, using Kubernetes Indexed Jobs to shard suites and Kueue quotas and priorities to manage shared capacity. It recommends balancing shards by historical duration, retrying a failed shard once on a fresh pod, and recording whether failures come from tests, the environment, or infrastructure so reporting can distinguish flakes and support release certification. Multiple clusters reduce blast radius but add routing, upgrades, and configuration overhead. The authors frame quotas, failure policy, and trustworthy reporting as core platform responsibilities, while noting centralization creates a shared dependency and is a poor fit for small organizations or workloads needing specialized hardware.

### Source excerpt

How a financial-services firm runs 250,000 automated tests a day on 4 Amazon EKS clusters with Kueue quotas, sharding, flake policy, and certification.

## Why Is My Kubernetes Job Still Pending? Tracing Kueue Admission with Headlamp

DevFeed: [Why Is My Kubernetes Job Still Pending? Tracing Kueue Admission with Headlamp](<https://devfeed.tech/articles/why-is-my-kubernetes-job-still-pending-tracing-kueue-admission-with-headlamp-87262.md>)

Original publisher: [Read original article](<https://kubernetes.io/blog/2026/10/14/tracing-kueue-admission-with-headlamp/>)

Author: Utkarsh Raj LFX Mentee; CNCF; René Dudfield; Kevin Hannon

Published: 2026-10-14T18:35:00Z

Content type: tutorial

Language: en

Sources: [Kubernetes Blog](<https://devfeed.tech/sources/kubernetes-blog.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>)

Tags: [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [queue](<https://devfeed.tech/tags/queue.md>), [quota](<https://devfeed.tech/tags/quota.md>), [scheduler](<https://devfeed.tech/tags/scheduler.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [tracing](<https://devfeed.tech/tags/tracing.md>)

### AI overview

A tutorial shows how to use the Headlamp Kueue plugin to trace why a Kubernetes Job remains suspended. It walks through Workload conditions, events, LocalQueue and ClusterQueue settings, quota, ResourceFlavors, and Cohorts, then explains when to continue troubleshooting in standard Job and Pod views. The 0.1.0-alpha plugin offers read-only views for Kueue resources.

### Source excerpt

You submit a Kubernetes Job and wait for its Pods to appear. A minute passes. Then another. The Job is still suspended, there are no application logs, and there may not even be a Pod to inspect. In a cluster managed by Kueue, this can be completely normal. Kueue may still be deciding whether the Job can use the requested resources. The useful question is not only "Why is the Pod pending?" It is first "Has Kueue admitted the Workload?" The Kueue plugin for Headlamp brings that investigation into the Headlamp UI. Instead of reconstructing the admission path from several commands and large YAML objects, you can begin with the Workload, read its condition message, and follow links through the LocalQueue, ClusterQueue, ResourceFlavor, and Cohort. This article follows that Headlamp-first path. The command line appears only in a short optional section for reproducing the example. Before you begin You need a Kubernetes cluster with Kueue installed, a Headlamp installation, and the 0.1.0-alpha Kueue plugin. Your Kubernetes identity must be allowed to read the Kueue resources used in this article. The alpha release provides read-only views for Kueue v1beta2 Cohorts, ClusterQueues, LocalQueues, ResourceFlavors, and Workloads. It does not add its own create, edit, stop, resume, or delete workflows. Headlamp still provides its standard YAML and resource actions where your permissions allow them. First, separate admission from Pod scheduling Kueue and the Kubernetes scheduler answer different questions. Kueue admission: Should this Workload start now, considering queue policy, quota, ResourceFlavors, admission checks, and other waiting Workloads? Pod scheduling: After the Workload is admitted and the Job resumes, which nodes can run its Pods? That boundary changes the investigation. If the Workload is not admitted, stay in the Kueue section of Headlamp. If it is admitted but its Pods are still Pending, move to Headlamp's normal Job and Pod views and inspect scheduler events. Kueu

## Kueue: A Kubernetes-Native System for AI Training Workloads

DevFeed: [Kueue: A Kubernetes-Native System for AI Training Workloads](<https://devfeed.tech/articles/kueue-a-kubernetes-native-system-for-ai-training-workloads-77991.md>)

Original publisher: [Read original article](<https://wf.coreweave.com/blog/kueue-a-kubernetes-native-system-for-ai-training-workloads>)

Author: CoreWeave

Published: 2026-10-02T14:12:41Z

Content type: tutorial

Language: en

Sources: [Coreweave Blog Posts](<https://devfeed.tech/sources/coreweave-blog-posts.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [GPU](<https://devfeed.tech/topics/gpu.md>)

Tags: [admission](<https://devfeed.tech/tags/admission.md>), [ai-ml](<https://devfeed.tech/tags/ai-ml.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [content-type](<https://devfeed.tech/tags/content-type.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>)

### AI overview

Kueue extends Kubernetes with queueing and workload admission controls for AI/ML batch jobs. It supports all-or-nothing and gang scheduling, resource flavors for matching workloads to hardware, and quota sharing among teams. CoreWeave Kubernetes Service integrates Kueue with a Helm chart and monitoring dashboard.

### Source excerpt

Discover how Kueue enhances Kubernetes for AI/ML training and inference workloads. Learn how CoreWeave integrates Kueue to optimize GPU scheduling and resource

## Microsoft Open-Sources TauGrid: A Kubernetes-Native Stack for GPU AI Workloads

DevFeed: [Microsoft Open-Sources TauGrid: A Kubernetes-Native Stack for GPU AI Workloads](<https://devfeed.tech/articles/microsoft-open-sources-taugrid-a-kubernetes-native-stack-for-gpu-ai-workloads-49861.md>)

Original publisher: [Read original article](<https://www.marktechpost.com/2026/09/17/microsoft-open-sources-taugrid-a-kubernetes-native-stack-for-gpu-ai-workloads/>)

Author: Asif Razzaq

Published: 2026-09-17T21:21:08Z

Content type: release

Language: en

Sources: [MarkTechPost](<https://devfeed.tech/sources/marktechpost.md>)

Topics: [AI Platform](<https://devfeed.tech/topics/ai-platform.md>), [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [Monitoring](<https://devfeed.tech/topics/monitoring.md>), [observability](<https://devfeed.tech/topics/observability.md>), [Command-line interface](<https://devfeed.tech/topics/cli.md>), [Helm charts](<https://devfeed.tech/topics/helm-charts.md>), [Go Language](<https://devfeed.tech/topics/go-language.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-shorts](<https://devfeed.tech/tags/ai-shorts.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [cli](<https://devfeed.tech/tags/cli.md>), [editors-pick](<https://devfeed.tech/tags/editors-pick.md>), [for-devs](<https://devfeed.tech/tags/for-devs.md>), [github](<https://devfeed.tech/tags/github.md>), [go](<https://devfeed.tech/tags/go.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [helm](<https://devfeed.tech/tags/helm.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kubernetes-native](<https://devfeed.tech/tags/kubernetes-native.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [observability](<https://devfeed.tech/tags/observability.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [python](<https://devfeed.tech/tags/python.md>), [staff](<https://devfeed.tech/tags/staff.md>), [tech-news](<https://devfeed.tech/tags/tech-news.md>), [technology](<https://devfeed.tech/tags/technology.md>)

### AI overview

Microsoft's AKS engineering team open-sourced TauGrid, a self-hosted Kubernetes platform for GPU-based AI workloads. It combines CLI-based submission, Kueue queueing and admission, Ray orchestration, GPU health monitoring, and workload observability in a Helm deployment. The post describes its MIT license, public container artifacts, supported Kubernetes prerequisites, workload recovery features, and evidence capture for reproducibility and auditing.

### Source excerpt

Microsoft's AKS engineering team open-sourced TauGrid on August 28, 2026, packaging the tau CLI, Kueue queueing, KubeRay orchestration, GPU node health monitoring and observability into one Helm install. It is MIT licensed and deployable now on any Kubernetes 1.30+ cluster with GPU nodes, kubectl and Helm 3.0 or later. The post Microsoft Open-Sources TauGrid: A Kubernetes-Native Stack for GPU AI Workloads appeared first on MarkTechPost.

## Microsoft Open-Sources TauGrid to Simplify AI Workload Management on Kubernetes

DevFeed: [Microsoft Open-Sources TauGrid to Simplify AI Workload Management on Kubernetes](<https://devfeed.tech/articles/microsoft-open-sources-taugrid-to-simplify-ai-workload-management-on-kubernetes-31518.md>)

Original publisher: [Read original article](<https://www.infoq.com/news/2026/09/microsoft-taugrid-open-source/>)

Author: Sergio De Simone

Published: 2026-09-16T18:00:00Z

Content type: news

Language: en

Sources: [InfoQ](<https://devfeed.tech/sources/infoq.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [Go Language](<https://devfeed.tech/topics/go-language.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [azure](<https://devfeed.tech/tags/azure.md>), [development](<https://devfeed.tech/tags/development.md>), [devops](<https://devfeed.tech/tags/devops.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-taugrid-open-source](<https://devfeed.tech/tags/microsoft-taugrid-open-source.md>), [news](<https://devfeed.tech/tags/news.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>)

### AI overview

Microsoft has open-sourced TauGrid, a cloud-native platform for managing, scheduling, and monitoring AI workloads on GPU-enabled Kubernetes clusters. It combines workload submission, Kueue-based queuing, KubeRay orchestration, GPU-node monitoring, and observability, while planned capabilities remain on its roadmap.

### Source excerpt

Microsoft has open-sourced TauGrid, a cloud-native platform designed to manage, schedule, and monitor AI workloads on GPU-enabled Kubernetes clusters. By Sergio De Simone

## Monitor TAS and gang scheduling for AI training in Kubernetes

DevFeed: [Monitor TAS and gang scheduling for AI training in Kubernetes](<https://devfeed.tech/articles/monitor-tas-and-gang-scheduling-for-ai-training-in-kubernetes-26969.md>)

Original publisher: [Read original article](<https://www.datadoghq.com/blog/monitor-tas-and-gang-scheduling-for-ai-training-in-kubernetes/>)

Author: David Lentz; Kathy Lin

Published: 2026-09-15T00:00:00Z

Content type: article

Language: en

Sources: [Datadog](<https://devfeed.tech/sources/datadog-the-monitor-blog.md>)

Topics: [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [Monitoring](<https://devfeed.tech/topics/monitoring.md>), [datadog](<https://devfeed.tech/topics/datadog.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [batch](<https://devfeed.tech/tags/batch.md>), [containers](<https://devfeed.tech/tags/containers.md>), [datadog](<https://devfeed.tech/tags/datadog.md>), [distributed-training](<https://devfeed.tech/tags/distributed-training.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-monitoring](<https://devfeed.tech/tags/gpu-monitoring.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [scheduler](<https://devfeed.tech/tags/scheduler.md>)

### AI overview

This article explains why Kubernetes scheduling is insufficient for distributed AI training workloads and how topology-aware scheduling and gang scheduling address hardware placement and simultaneous startup requirements. It discusses implementing these capabilities with Kueue and the Coscheduling plugin, and monitoring and troubleshooting them with Datadog GPU Monitoring.

### Source excerpt

Learn how Datadog helps you correlate Kueue, Coscheduling, GPU, and training framework signals to validate gang scheduling and topology-aware scheduling.

## China Merchants Bank Wins CNCF End User Case Study Contest for Unifying AI Training and Inference on Kubernetes

DevFeed: [China Merchants Bank Wins CNCF End User Case Study Contest for Unifying AI Training and Inference on Kubernetes](<https://devfeed.tech/articles/china-merchants-bank-wins-cncf-end-user-case-study-contest-for-unifying-ai-training-and-inference-on-kubernetes-4594.md>)

Original publisher: [Read original article](<https://www.cncf.io/announcements/2026/09/07/china-merchants-bank-wins-cncf-end-user-case-study-contest-for-unifying-ai-training-and-inference-on-kubernetes/>)

Author: Haley White

Published: 2026-09-08T01:54:31Z

Content type: news

Language: en

Sources: [Cloud Native Computing Foundation](<https://devfeed.tech/sources/cloud-native-computing-foundation.md>)

Topics: [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [AI Platform](<https://devfeed.tech/topics/ai-platform.md>), [Inference](<https://devfeed.tech/topics/inference.md>), [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Cloud Native Ecosystem](<https://devfeed.tech/topics/cloud-native-ecosystem.md>), [data centers](<https://devfeed.tech/topics/data-centers.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [announcements](<https://devfeed.tech/tags/announcements.md>), [case-study](<https://devfeed.tech/tags/case-study.md>), [china](<https://devfeed.tech/tags/china.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cost](<https://devfeed.tech/tags/cost.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [inference](<https://devfeed.tech/tags/inference.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [lora](<https://devfeed.tech/tags/lora.md>)

### AI overview

China Merchants Bank won a CNCF case-study contest for a Kubernetes-based AI platform that shares nearly 10,000 accelerator cards across training, fine-tuning, and online inference. The bank reports increased average accelerator utilization and lower inference costs.

### Source excerpt

New cloud native platform lifted average accelerator compute utilization from 35% to more than 60% and cut inference cost per 1 million tokens by more than 60% Key Highlights SHANGHAI, China - KubeCon + CloudNativeCon +...

## How Netflix Simplified Batch Compute with Kueue

DevFeed: [How Netflix Simplified Batch Compute with Kueue](<https://devfeed.tech/articles/how-netflix-simplified-batch-compute-with-kueue-139.md>)

Original publisher: [Read original article](<https://netflixtechblog.com/how-netflix-simplified-batch-compute-with-kueue-87860682629c?source=rss----2615bd06b42e---4>)

Author: Netflix Technology Blog

Published: 2026-06-22T21:35:01Z

Content type: article

Language: en

Sources: [Netflix](<https://devfeed.tech/sources/netflix-techblog-medium.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>)

Tags: [aws](<https://devfeed.tech/tags/aws.md>), [batch](<https://devfeed.tech/tags/batch.md>), [compute](<https://devfeed.tech/tags/compute.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kubernetes-native](<https://devfeed.tech/tags/kubernetes-native.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [migration](<https://devfeed.tech/tags/migration.md>), [netflix](<https://devfeed.tech/tags/netflix.md>), [queuing](<https://devfeed.tech/tags/queuing.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>)

### AI overview

Netflix moved CMB's queueing and scheduling to Kueue while keeping its existing user-facing API and Titus scheduler profiles. Titus federation routes jobs to Kueue-enabled cells; tenant enrollment maps internal tenants to Cohorts and leaf tenants to ClusterQueue and LocalQueue, with capacity settings represented as resource flavors and nominal quotas. Netflix chose Kueue over options such as YuniKorn and Volcano because it leaves pod scheduling to kube-scheduler, supports multi-tenant quotas across heterogeneous hardware, and includes features such as preemption and topology-aware scheduling. The team migrated its largest, most complex customer early; production migration then took four weeks. Load testing also revealed that Kueue needed higher QPS, burst, and groupKindConcurrency settings than defaults. Netflix reports Kueue now manages millions of batch workloads. Preemption-based fair sharing lets tenants borrow idle reserved capacity and allows higher-priority jobs to preempt lower-priority ones; the excerpt reports a significant increase in average resource utilization.

### Source excerpt

By Alvin Bao, Alex Petrov, Jennifer Lai, Aidan Sherr, and Samartha Chandrashekar As a part of the journey to transition Netflix's compute infrastructure to be more Kubernetes-native, we have leaned into incorporating components from the Kubernetes ecosystem into our container platform Titus. One example of this is our use of Kueue, a cloud-native job queueing system for batch workloads, which has largely replaced the custom queuing and scheduling logic in our homegrown managed batch solution Compute Managed Batch (CMB). In this post, we'll give an overview of what motivated the migration, how we migrated millions of batch jobs to use Kueue, and what Kueue allows us to offer as a Compute platform. Brief Overview of CMB and Titus CMB is a managed batch solution that allows users and applications to execute and manage workloads that run to completion. Using a tenant hierarchy, workloads are managed and queued with ordered execution through priorities, and capacity is managed on a per-tenant basis. Workloads that are submitted to CMB are then run on Titus. The features of Titus relevant to CMB are workload federation across multiple cells (Kubernetes clusters) and federated capacity reservations. This means CMB can talk to a single Titus endpoint to get/submit workloads and update capacity reservations without having to worry about the underlying cell/cluster topology. CMB Tenant Hierarchy Tenants provide a grouping mechanism for jobs submitted on behalf of certain organizations, platforms, or applications. Users can create and organize tenants however best suits their organization or use case. For example, an organization may use a single tenant across several applications or a complex hierarchical structure that matches its team and application ownership structure. Tenants are associated with a capacity configuration. The capacity configuration defines the amount of compute capacity available to the tenant and provides certain guarantees around isolation from other te

## Kubernetes v1.36: Mutable Pod Resources for Suspended Jobs (beta)

DevFeed: [Kubernetes v1.36: Mutable Pod Resources for Suspended Jobs (beta)](<https://devfeed.tech/articles/kubernetes-v1-36-mutable-pod-resources-for-suspended-jobs-beta-4540.md>)

Original publisher: [Read original article](<https://kubernetes.io/blog/2026/04/27/kubernetes-v1-36-mutable-pod-resources-for-suspended-jobs/>)

Author: Kevin Hannon

Published: 2026-04-27T18:35:00Z

Content type: article

Language: en

Sources: [Kubernetes Blog](<https://devfeed.tech/sources/kubernetes-blog.md>)

Topics: [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [jobs](<https://devfeed.tech/topics/jobs.md>), [kueue](<https://devfeed.tech/topics/kueue.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [cpu](<https://devfeed.tech/topics/cpu.md>), [Availability](<https://devfeed.tech/topics/availability.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [availability](<https://devfeed.tech/tags/availability.md>), [batch](<https://devfeed.tech/tags/batch.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [jobs](<https://devfeed.tech/tags/jobs.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kueue](<https://devfeed.tech/tags/kueue.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>)

### AI overview

Kubernetes v1.36 promotes mutable CPU, memory, GPU, and extended-resource requests and limits for suspended Jobs to beta. Queue controllers such as Kueue can adjust resources before a Job resumes, avoiding deletion and recreation when cluster capacity or workload requirements change.

### Source excerpt

Kubernetes v1.36 promotes the ability to modify container resource requests and limits in the pod template of a suspended Job to beta. First introduced as alpha in v1.35, this feature allows queue controllers and cluster administrators to adjust CPU, memory, GPU, and extended resource specifications on a Job while it is suspended, before it starts or resumes running. Why mutable pod resources for suspended Jobs? Batch and machine learning workloads often have resource requirements that are not precisely known at Job creation time. The optimal resource allocation depends on current cluster capacity, queue priorities, and the availability of specialized hardware like GPUs. Before this feature, resource requirements in a Job's pod template were immutable once set. If a queue controller like Kueue determined that a suspended Job should run with different resources, the only option was to delete and recreate the Job, losing any associated metadata, status, or history. This feature also provides a way to let a specific Job instance for a CronJob progress slowly with reduced resources, rather than outright failing to run if the cluster is heavily loaded. Consider a machine learning training Job initially requesting 4 GPUs: apiVersion: batch/v1 kind: Job metadata: name: training-job-example-abcd123 labels: app.kubernetes.io/name: trainer spec: suspend: true template: metadata: annotations: kubernetes.io/description: "ML training, ID abcd123" spec: containers: - name: trainer image: example-registry.example.com/training:2026-04-23T150405.678 resources: requests: cpu: "8" memory: "32Gi" example-hardware-vendor.com/gpu: "4" limits: cpu: "8" memory: "32Gi" example-hardware-vendor.com/gpu: "4" restartPolicy: Never A queue controller managing cluster resources might determine that only 2 GPUs are available. With this feature, the controller can update the Job's resource requests before resuming it: apiVersion: batch/v1 kind: Job metadata: name: training-job-example-abcd123 labels

## Guide to Kubernetes Resource Quota: Examples & Pros and Cons

DevFeed: [Guide to Kubernetes Resource Quota: Examples & Pros and Cons](<https://devfeed.tech/articles/guide-to-kubernetes-resource-quota-examples-pros-and-cons-80052.md>)

Original publisher: [Read original article](<https://www.groundcover.com/blog/kubernetes-resource-quota>)

Author: Aviv Zohari

Published: 2025-03-20T00:00:00Z

Content type: tutorial

Language: en

Sources: [Ground Cover](<https://devfeed.tech/sources/ground-cover.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [The Three Main Cloud Service Models](<https://devfeed.tech/topics/the-three-main-cloud-service-models.md>)

Tags: [best-practices](<https://devfeed.tech/tags/best-practices.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [examples](<https://devfeed.tech/tags/examples.md>), [guide](<https://devfeed.tech/tags/guide.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [memory](<https://devfeed.tech/tags/memory.md>), [quota](<https://devfeed.tech/tags/quota.md>), [storage](<https://devfeed.tech/tags/storage.md>)

### AI overview

This tutorial explains how Kubernetes resource quotas limit resource requests, consumption and object counts within namespaces. It compares quotas with limit ranges and requests, gives configuration examples, and discusses tradeoffs and practices for managing CPU, memory, storage and other resources.

### Source excerpt

Learn how to use Kubernetes resource quotas to manage CPU, memory & storage. Explore examples, pros, cons, and best practices.

## How we minimized the overhead of Kubernetes in our job system

DevFeed: [How we minimized the overhead of Kubernetes in our job system](<https://devfeed.tech/articles/how-we-minimized-the-overhead-of-kubernetes-in-our-job-system-83482.md>)

Original publisher: [Read original article](<https://www.datadoghq.com/blog/engineering/moving-a-jobsystem-to-kubernetes/>)

Author: Lally Singh; Ashwin Venkatesan

Published: 2021-02-22T00:00:00Z

Content type: article

Language: en

Sources: [Datadog](<https://devfeed.tech/sources/datadog.md>)

Topics: [kueue](<https://devfeed.tech/topics/kueue.md>), [Kernel](<https://devfeed.tech/topics/kernel.md>), [Deployment-Driven Development](<https://devfeed.tech/topics/deployment-driven-development.md>)

Tags: [clusters](<https://devfeed.tech/tags/clusters.md>), [containers](<https://devfeed.tech/tags/containers.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [delay](<https://devfeed.tech/tags/delay.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [performance](<https://devfeed.tech/tags/performance.md>)

### AI overview

Datadog describes moving its job system to Kubernetes and diagnosing an initial performance regression. By comparing equivalent workloads, the team found that pod resource requests limited scheduling density and that slow worker status checks left CPUs idle. Adjusting the requests and reducing the check interval cut the cluster from 200 nodes to 138; the post estimates that further tuning could bring it to about 120, while noting kernel and CPU mitigation differences from the VM baseline.

### Source excerpt

Solving performance problems when moving an application to Kubernetes