Skip to main content

Choosing a Hosting Topology

How you host CMP itself (POC vs production vs HA) is separate from the logical CMP architecture. Use this page for capacity planning before you provision VMs.

CloudStack’s own guide for cloud infrastructure scale is Choosing a Deployment Architecture — that document is about zones/pods, not CMP VMs.

TopologyTypical useFootprint
Single serverPOC, staging1 VM — all roles co-located
Multi-serverStandard productionSeparate frontend, backend, database
HA multi-tierLarge production18 servers — redundant web, proxy, app, cache, database
tip

Validate on single server, go live on multi-server, adopt HA when you need redundant load balancers and database failover.

Open full size

Click a diagram, or use Open full size, to view it in a new browser tab.


Single-server deployment

One host runs NGINX, frontend, backend, PostgreSQL, Redis, and the scheduler.

Diagram: CMP single-server deployment
Open full size
TopicBehaviour
ComplexityLowest
AvailabilitySingle point of failure
SpecsStaging / Single VM — full requirements
warning

Do not keep a busy production cloud on single-server long term.


Multi-server deployment

Standard production split:

VMRole
FrontendNGINX + Customer Portal
BackendCMP API, workers, scheduler
DatabasePostgreSQL (Redis often on backend or dedicated cache)
Diagram: CMP single-server deployment
Open full size

See Production — three-VM requirements.


HA multi-tier deployment

Redundant edge, web, proxy, application, Redis, DB proxy, and database tiers. A full HA layout requires 18 servers in total.

important

Per-role sizing and the exact server breakdown are confirmed with the StackConsole team. See HA — Server Requirements.

img/architecture/cmp-ha-multi-server.png

Diagram: CMP HA / multi-tier deployment
Open full size
TierNotes
EdgeCloudflare / Akamai-style global LB
WebMultiple NGINX + portal nodes (public subnet)
ProxyHAProxy / Keepalived pairs
ApplicationMultiple API / worker nodes (private subnet), shared filesystem
CacheRedis cluster
DatabasePrimary + replicas behind HA DB proxies
important

HA requires ownership of TLS, health checks, shared storage, Redis topology, PostgreSQL failover, and backup drills. Engage StackConsole when designing this layout for the first time.


How to choose

QuestionPrefer
First install / training?Single server
Live customers, moderate load?Multi-server
Strict uptime / high traffic?HA multi-tier

For the logical CMP runtime view (load balancer, API, providers), see CMP Deployment Architecture.


Next steps