Arcfra Kubernetes Engine (AKE for short) is a Kubernetes service provided by the Arcfra enterprise cloud platform. Based on Arcfra virtualization, AKE can automatically create multiple virtual machines to build highly available Kubernetes clusters and can also use physical machines as worker nodes. It supports NVIDIA GPUs to provide parallel computing capabilities for clusters, including AI inference and training, as well as high-performance computing. With Arcfra production-grade distributed storage and CSI (Container Storage Interface) addons, AKE provides stable, high-performance persistent volumes for stateful applications. With Arcfra network and security products and CNI (Container Network Interface) addons, AKE enables flat interconnection between virtual machines and containers, along with unified network security policy management.
AKE pre-integrates addons for service exposure, application deployment, monitoring, logging, and other capabilities in the cloud-native ecosystem, shortening the time required for Kubernetes clusters to become ready. AKE also supports other applications in the Kubernetes ecosystem. Through a unified graphical interface, O&M personnel can manage the full lifecycle of all clusters and container applications, and view their monitoring, logging, event, and audit information.
With these capabilities, enterprise O&M teams can easily deploy, manage, and use production-grade Kubernetes clusters on servers with multiple CPU architectures, reducing the difficulty of enterprise cloud-native transformation and accelerating application modernization.
Technical features
- Convenient and efficient K8s cluster lifecycle management
- Implements a declarative K8s cluster orchestration engine based on Cluster API, a mature open-source project from the K8s community.
- Deeply integrates with the ACOS hyperconverged architecture to quickly build production-grade K8s clusters that are highly available and easy to operate and maintain. It also supports ACOS active-active clusters to provide high availability and service continuity across data centers.
- The self-developed Cluster API provider supports diverse infrastructure environments, with VM and physical-machine node types, and multiple operating systems and K8s versions available.
- Integrates NVIDIA GPU Operator to provide both device passthrough and vGPU time-slicing modes.
- Provides auto scaling and automatic replacement of faulty nodes based on node groups to implement cluster elasticity and high availability.
- Implements in-place updates of cluster and node configurations based on Arcfra virtualization and the node configuration proxy addon.
- Connects to the observability platform to provide unified monitoring and alerting capabilities.
- Scalable cluster addon management
- Supports multiple common K8s infrastructure services as addons.
- Supports enabling or disabling addons with one click, and allows addon parameters to be configured as needed.
- High-performance persistent storage for K8s containers
- Directly mounts ACOS virtual disks as persistent volumes (PVs) through the AVE CSI driver.
- Connects to ABS iSCSI LUNs through the ABS CSI driver to provide persistent volumes (PVs).
- Secure and reliable Arcfra Integrated CNI
- Implements a flat container network, breaking the boundary between container networks and VM networks.
- Coordinates container network security policies with virtualization microsegmentation services.
- Simple and easy-to-use visual management
- Provides a unified graphical management interface based on AOC to manage clusters, nodes, and container applications on a single platform.
- Lowers the technical threshold for application launch and O&M through form-based deployment and management of K8s resources.
- Provides a unified topology view of application-related resources, visually presenting application architecture and resource relationships.
- Provides convenient access entries for clusters, nodes, and Pods through the CloudShell web terminal proxy.
- Fine-grained multi-tenancy and permission management
- Uses RBAC permission control and organizes namespaces by project to implement tenant-level isolation of resource access policies.
- Implements identity authentication for cluster access based on the webhook token authentication mechanism of the K8s TokenReview API.
- Extends and optimizes the native K8s resource quota mechanism to support fine-grained resource management for project-level and namespace-level quotas.
- Implements tenant-level Kubernetes resource access control based on the user permission model and API gateway proxy.
