- Standard
- Control plane managed by AWS including storage and scaling
- Data layer managed by user
- Full control over underlying architecture of workers
- More responsibility, such as scaling and lifecyle automation
- Auto Mode
- Control plane and worker nodes managed by AWS
- Manages cluster and day-to-day operations
- More streamlined, less responsibility
- Ideal for smaller teams or teams with minimal Kubernetes knowledge
- Fewer features and customization
- Manage state of the overall infrastructure
- The API Server (kube-apiserver): communication between clusters
- etcd - distributed key-value storage to store state of clusters
- Scheduler: pods -> nodes
- Kubenetes Controller Manager: keep components in their desired state
- Managed Node Groups: Provision and managed by EKS: updating, health, replacement, provisioned as part of auto scaling groups, can use customized launch templates
- Self Managed: node groups provisioned by the user, allows for customization of the underling system, can use auto scaling groups
- Serverless worker nodes
- Handles: server type, auto scale, other configuration
- User only need to define Fargate profile, namespace, execution role








EKS Communication
EKS Permission