CloudStack organizes infrastructure hierarchically. Each level serves a specific purpose.

The Hierarchy

Region (Geographically dispersed)
  └─ Zone (Usually = 1 datacenter)
      └─ Pod (Usually = 1 physical rack)
          └─ Cluster (Same hypervisor type)
              └─ Host (Individual physical computer)

Regions

Collection of zones across different locations. If one datacenter fails, other regions keep running.

Visible to users: Yes

Zones

Typically one datacenter. Provides physical isolation (separate power, network). Contains pods and secondary storage.

Visible to users: Yes

Pods

Usually one physical rack. Contains clusters and primary storage. All hosts in a pod are on the same subnet.

Visible to users: No

Clusters

Identical hosts running the same hypervisor. Enables live migration between hosts without downtime.

Requirements:

  • Same hypervisor (all KVM, or all Hyper-V, etc.)
  • Identical hardware
  • Same subnet
  • Shared Primary Storage

Visible to users: No

Hosts

Individual physical computers with hypervisor software. Run the instances.

Visible to users: No

Storage

Primary Storage: Cluster/zone-level. Stores virtual disks of running instances.

Secondary Storage: Zone-wide. Stores templates, ISOs, snapshots.

Networking

Basic: Single network per pod. Layer-3 isolation (firewall rules). Not for multi-tenant.

Advanced: Multiple networks per zone. Layer-2 isolation (VLANs). Suitable for multi-tenant.

Traffic types:

  • Guest: Instance-to-instance
  • Management: CloudStack system communication
  • Storage: Template/snapshot copying
  • Public (advanced only): Internet access

Why This Matters

Each level isolates failures:

  • Region failure: Other regions survive
  • Zone failure: Other zones survive
  • Pod failure: Other pods survive
  • Cluster failure: Other clusters survive

Homogeneous clusters enable live migration (zero-downtime maintenance).

Different hypervisors in different clusters gives flexibility.