Comparing Managed Kubernetes clusters on dedicated and cloud servers

	Cloud server	Dedicated server
What tasks is it suitable for	For tasks with uneven load when node group autoscaling is needed	For tasks with stable load and for stream data processing
Resource usage	Multiple clients can be on the same host: if one client uses all host resources, it affects the performance of other clients on that host; due to a shared bandwidth, DDoS attacks on one client can affect other clients	Dedicated resources and isolation from other clients
Payment	Pay-as-you-go — hourly billing only for used resources	Billing options with prepayment for all ordered resources
Data storage	Persistent volumes (PVs) on network volumes. If a node fails, the network volume will be preserved and no data will be lost	local storage (for example, local disks, persistent volumes (PVs) on the same server); external storage (for example, file storage, persistent volumes (PVs) on another server)
Node group autoscaling	✓	✗
Changing the number of nodes in a group	✓	✓
Changing node configuration	✓	✗
Configurations with GPU	✓	✓
Deployment time	3-5 minutes	Up to 60 minutes
Private Kube API	✓	✓
Cluster certificate update	✓	✓
Integration with Container Registry	✓	✓
Minor version update	✓	✗
Patch version auto-update	✓	✗
Node autorepair	✓	✗
Labels	✓	✓
Taints	✓	✓
User data	✓	✗
Node reboot	3-5 minutes	5-10 minutes
Node reinstallation	✓	✓
Terraform support	✓	✗