#Infra #IaaS #Concepts
| Term | General Infrastructure Perspective | On-Premises (OpenStack/KVM) | Cloud (AWS/EKS/ECS) | Workload Examples |
|---|---|---|---|---|
| Server | Physical hardware with CPU, RAM, Disk, and NIC | Bare-metal servers (Dell/HPE, etc.) | AWS EC2 instance (virtual server) | The hardware itself — no apps running yet |
| Node | A server unit assigned a specific role | Controller Node, Compute Node | EKS Worker Node, ECS Task, Control Plane node | Designating a role like “this server is Compute-only” |
| Controller | The “brain” of the cluster — manages and orchestrates | Keystone, Glance, Neutron-Server, Scheduler | AWS EKS/ECS Control Plane (managed by AWS) | Doesn’t run apps directly. Instead, issues directives like “run this workload on Compute node 3” |
| Compute | The actual workload executor | OpenStack Nova-Compute (runs VMs) | EC2 instances, ECS Tasks, EKS Worker Nodes | Django web servers, MySQL DBs, AI inference, log analysis, etc. |
| Storage Node | Stores, replicates, and serves data | Ceph OSD, Swift/Object Storage | EBS, EFS, FSx, S3 | DB data files, user-uploaded images, logs |
| Network Node (Gateway) | Forwards and routes traffic | Neutron L3/NAT, DHCP Agent | VPC, Subnet, NAT GW, ALB/NLB | Routing external user requests to web servers |
| Availability Zone (AZ) | Physical isolation to prevent fault propagation | Separate racks, power, and switches | AWS AZ (ap-northeast-2a, 2b, 2c) | If one AZ goes down, workloads continue running in other AZs |