Overview

Introduction

Cluster federation is a multi-cloud container orchestration capability provided by Karmada. Cluster federation aims to manage multi-cluster applications in cross-cloud and cross-region scenarios, with features such as unified multi-cluster management, application deployment, service discovery, auto scaling, and failover.

Figure 1 Cluster federation architecture

Advantages

Cluster federation is fully compatible with native Kubernetes APIs and can seamlessly integrate with the existing Kubernetes tool chain ecosystem. With cluster federation, you can add more clusters to deploy your applications without reconstructing the applications.
Diverse multi-cluster scheduling policies: There are cluster affinity scheduling, multi-cluster splitting/rebalancing scheduling, and multi-dimensional HA deployment (including multi-region, multi-AZ, multi-cluster, and multi-cloud deployment). There is also automated graceful migration. When a cluster becomes faulty, service instances in this cluster are automatically migrated to a healthy cluster in a centralized or distributed manner to ensure that there are always service instances.
Multi-cluster traffic distribution: MCS provides cross-cluster service discovery and access. MCI provides cross-cluster load balancing and traffic routing. MCI also supports automatic traffic switchover. Traffic on a faulty cluster can be automatically redirected to a health cluster for service availability.
Multi-cluster auto scaling: Workloads in multiple clusters can be scaled based on system metric changes and custom metric changes or at regular intervals, improving the availability and stability of workloads.
Intelligent container analysis: Container Intelligent Analysis monitors applications and resources in real time and collects metrics and events for application statuses analysis. It provides unified full-stack monitoring views and end-to-end tracing and data visualization. It also provides cluster health diagnosis, greatly speeding up fault analysis and locating.