What Is a CCE Autopilot Cluster?

Introduction

CCE Autopilot allows you to create serverless clusters that offer optimized Kubernetes compatibility and free you from O&M. After a CCE Autopilot cluster is created, you can deploy applications without purchasing nodes or maintaining the deployment, management, and security of nodes. You only need to focus on the implementation of application service logic, which greatly reduces your O&M costs and improves the reliability and scalability of applications.

Advantages

Out-of-the-box usability: Applications can be directly deployed after Kubernetes clusters are created.
Focus on applications: You can focus on building and deploying applications without managing the underlying infrastructure of the clusters.
Security assurance: If there are available security patches, the clusters will be automatically upgraded as scheduled.
Lower costs: Applications are created and charged on demand, and no idle resources are produced, reducing the costs.
Node hosting: Worker nodes are hosted on CCE. You do not need to manually manage and upgrade these nodes.
Elastic pre-warming: When an HPA policy is configured for a workload, CCE automatically pre-warms new nodes for new pods to easily handle traffic spikes.
Appropriate scheduling: CCE Autopilot manages resource bin packing to improve resource utilization. You do not need to consider the number of pods running on each node.
Resource allocation: If resource requests (such as CPU request and memory request) are not specified for workload deployment, CCE Autopilot automatically allocates the required resources for the pods.
Simplified O&M: CCE Autopilot frees you from paying attention to node status. Cluster nodes can be automatically scaled in or out based on the load scale, which reduces the platform O&M burden.

Comparison Between CCE Autopilot, CCE Standard, and CCE Turbo

**Table 1** Comparison between different types of CCE clusters
Dimension	CCE Autopilot	CCE Standard	CCE Turbo
Node management	Worker nodes are fully managed. CCE Autopilot takes care of node scaling and pre-warming.	You need to take care of the management and O&M of worker nodes.	You need to take care of the management and O&M of worker nodes.
Node OSs	There are dedicated OSs that use containerd as the container engine.	You can select an OS and container engine.	You can select an OS and container engine.
Node specifications	Node specifications are adaptive to the workload scale.	You can select the node specifications as needed.	You can select the node specifications as needed.
Node upgrade and maintenance	Nodes are upgraded and recovered automatically.	Nodes need to be reset for upgrade.	Nodes need to be reset for upgrade.
Container network model	Cloud native 2.0 network	VPC network Tunnel network	Cloud native 2.0 network
Network performance	The VPC network and container network are flattened into one for zero performance loss.	The container network is overlaid with the VPC network, causing performance loss.	The VPC network and container network are flattened into one for zero performance loss.
Network isolation	Pods can be associated with security groups for isolation.	Tunnel network model: network policies for communications within a cluster VPC network model: isolation not supported	Pods can be associated with security groups for isolation.
Services	ClusterIP LoadBalancer	ClusterIP NodePort LoadBalancer DNAT	ClusterIP NodePort LoadBalancer DNAT
Ingresses	ELB Ingress Nginx Ingress	ELB Ingress Nginx Ingress	ELB Ingress Nginx Ingress
Add-ons	The following configurable add-ons are available: CoreDNS Kubernetes Metrics Server CCE Advanced HPA Cloud Native Cluster Monitoring Cloud Native Logging NGINX Ingress Controller	The following configurable add-ons are available: CoreDNS CCE Container Storage (Everest) CCE Node Problem Detector Kubernetes Dashboard CCE Cluster Autoscaler Kubernetes Metrics Server CCE Advanced HPA CCE Cloud Bursting Engine for CCI CCE AI Suite (NVIDIA GPU) CCE AI Suite (Ascend NPU) Volcano Scheduler NGINX Ingress Controller CCE Secrets Manager for DEW CCE Network Metrics Exporter NodeLocal DNSCache Cloud Native Cluster Monitoring Cloud Native Logging	The following configurable add-ons are available: CoreDNS CCE Container Storage (Everest) CCE Node Problem Detector Kubernetes Dashboard CCE Cluster Autoscaler Kubernetes Metrics Server CCE Advanced HPA CCE Cloud Bursting Engine for CCI CCE AI Suite (NVIDIA GPU) CCE AI Suite (Ascend NPU) Volcano Scheduler NGINX Ingress Controller CCE Secrets Manager for DEW CCE Network Metrics Exporter NodeLocal DNSCache Cloud Native Cluster Monitoring Cloud Native Logging
Local storage	emptyDir	Local PV hostPath emptyDir	Local PV hostPath emptyDir
Cloud storage	SFS SFS Turbo OBS	EVS SFS SFS Turbo OBS DSS	EVS SFS SFS Turbo OBS DSS

Constraints

Worker nodes in CCE Autopilot clusters are fully hosted. For this reason, some features that depend on the node features, such as hostPath and hostNetwork, are not supported.

Unavailable Feature	Description	Recommended Alternative Solution
DaemonSets	Deploys pods on each node.	Deploy multiple images in a pod using sidecars.
Setting hostPath in a pod	Mounts local files of a node to a container.	Use emptyDir or cloud storage of any type.
Setting hostNetwork in a pod	Maps a node port to a container port.	Use Services of the LoadBalancer type.
Use Services of the NodePort type.	Opens a node port to access containers.	Use Services of the LoadBalancer type.

When a CCE Autopilot cluster is used, the container storage space on the backend instance is 20 GB so you should use container images less than or equal to 5 GB. In addition, if a large number of files are created in the container root directory or an emptyDir volume when a container is running normally, you should use external storage (such as SFS, SFS Turbo, and OBS).
If a CCE Autopilot cluster is used, a maximum of 500 pods can be created. Add-on pods may occupy the pod quota. Plan the pod quota properly.
If a CCE Autopilot cluster is used, workloads that use Arm images are not supported.