DEV Community


Posted on

KubeSphere 3.2.0 GA: Bringing AI-oriented GPU Scheduling and Flexible Gateways to Kubernetes Management

KubeSphere, a fast-growing open source community, today announced the general availability of KubeSphere 3.2.0, the latest version of a comprehensive Kubernetes management platform. 

Monitoring GPU usage and management of GPU resource quotas are essential in the era of artificial intelligence (AI) and machine learning. With KubeSphere 3.2.0, users can create GPU workloads on the GUI, schedule GPU resources, and manage GPU resources quotas by tenants. Moreover, it is compatible with NVIDIA and vGPU solutions.

Six months ago, KubeSphere introduced edge computing features, metering and billing in its v3.1.0, extending Kubernetes from the cloud to the edge and helping enterprises more clearly assess the operating cost of infrastructure and applications. Now with the general availability of KubeSphere 3.2.0, enhanced features such as multiple Kubernetes cluster management, multi-tenant management, observability, DevOps, App store, and service mesh further perfect interactive design for better user experience.

CNCF Survey 2020 shows that over 80% of users run more than two Kubernetes clusters in their production environment. KubeSphere aims at addressing multi-cluster and multi-cloud challenges. It provides a unified control plane and supports distributing applications and replicas to multiple clusters deployed across public cloud and on-premises environments. Moreover, KubeSphere supports observability across clusters, including monitoring, logs, events, and auditing logs in multiple dimensions.

KubeSphere stays committed to open source with a vibrant ecosystem, which boasts a variety of lightweight, efficient toolkits to make Kubernetes easy. KubeKey, an open source installer, uses Docker as the default container runtime to install Kubernetes efficiently, and now it also supports CRI runtimes like containerd, CRI-O, and iSula after Dockershim deprecation. Other smart tools such as OpenFunction, OpenELB (CNCF Sandbox), Fluent Operator, are all open to everybody and users have access to code and documentation in GitHub repositories.

As an open source project, KubeSphere is making great strides on its international journey with over 700,000 downloads across 90 countries and regions to date. Not limited to simplifying the learning curve of Kubernetes for individual users, KubeSphere also gains significant popularity among medium and large enterprises for its competency in security, scalability, high availability, and more. In 2021, KubeSphere announced the collaboration with Amazon Web Services (AWS)Quick Start, DigitalOcean Marketplace, and Microsoft Azure Marketplace, further accelerating its adoption and internationalization in the booming cloud-native industry.

“Open source and internationalization are two solid pillars to KubeSphere, ” said Ray Zhou, Head of the KubeSphere project. “Tens of thousands of individual users and enterprises across the globe are using KubeSphere in production environments, and what we can repay them is to continue our commitment to providing cutting-edge cloud-native stacks and helping enterprises to embrace Kubernetes with ease.”

About KubeSphere

KubeSphere is a distributed operating system for cloud-native application management, using Kubernetes as its kernel. It provides a plug-and-play architecture, allowing third-party applications to be seamlessly integrated into its ecosystem.
KubeSphere offers wizard web UI and various enterprise-grade features for operation and maintenance, including multi-cloud and multi-cluster management, DevOps (CI/CD), application lifecycle management, service mesh, multi-tenancy, observability, storage and network management, and GPU support. 
For more information, you can visit or

Top comments (0)