Mental model
Cisco objective 4.2 says “Describe the attributes of different application deployment models (private cloud, public cloud, hybrid cloud, and edge)”. Four buckets.
| Model | Who owns the DC | Who runs the DC | Where it lives | Cost model |
|---|---|---|---|---|
| Private cloud | You | You | Your facility (or your colo) | Capex-heavy + staff |
| Public cloud | Provider (AWS/Azure/GCP/Cisco+) | Provider | Provider’s DCs globally | Pay per use (opex) |
| Hybrid cloud | Mixed | Mixed | Both, with network between | Mixed |
| Edge | You (usually) | You | Many small sites (branch / factory / tower) | Many small boxes |
Private cloud
- You bought the hardware. You staff the DC. You eat the electricity bill.
- Benefits: total control, data stays in your facility, no per-call API cost.
- Downsides: high up-front investment, you absorb demand spikes.
- Platforms: vSphere, OpenStack, Cisco UCS + ACI, Nutanix.
Public cloud
- Provider owns everything. You consume by API.
- Benefits: elastic (spin up 100 servers for 1 hour, pay for 100 server-hours), no hardware refresh, broad feature catalogue.
- Downsides: variable OpEx, data egress fees, lock-in risk, less control over hardware or network.
- Big names: AWS, Azure, GCP, Oracle Cloud; Cisco Intersight is a SaaS cloud too.
Hybrid cloud
- Workloads live in both private and public. Data flows between them.
- Common patterns:
- Burst to cloud — baseline capacity on-prem; overflow to public cloud during peaks.
- Dev/test in cloud, prod on-prem — cheap experimentation.
- Steady-state on-prem, DR in cloud — disaster recovery to a hot standby in the cloud.
- Containerized app on both — Kubernetes clusters in each; the app knows where it is.
Edge
- Many small compute sites near where data is produced or where users live.
- Not a competing model so much as a complement: data is pre-processed at the edge, summaries go to a central cloud.
- See the edge computing topic for the full why.
Comparison table
| Dimension | Private | Public | Hybrid | Edge |
|---|---|---|---|---|
| Capital investment | High | None | Mixed | Medium (many small boxes) |
| Operating cost shape | Steady | Variable | Mixed | Steady per site |
| Elasticity | Limited | Essentially infinite | Varies | Per-site |
| Latency to end user | DC-dependent | Regional | Mixed | Lowest |
| Vendor lock-in risk | Low | High | Medium | Low |
| Compliance / data sovereignty | Easy | Complex | Complex | Easy (data stays local) |
When to pick which
- Private when regulatory or performance requires total control and the workload is predictable.
- Public when you need to scale up fast or geographically, or when CapEx is unavailable.
- Hybrid when some workloads are sensitive (DBs with PII) but others benefit from elasticity (web front end).
- Edge when latency, local survival, or bandwidth to central cost demand it.
FAQ
Is hybrid cloud always better? No. Hybrid is complex: two environments, two sets of tooling, double the auth. Only choose it when you have a reason that justifies the complexity.
Is “multi-cloud” the same as “hybrid cloud”? Different. Multi-cloud = multiple public clouds (AWS + Azure). Hybrid = at least one private + at least one public. Both can coexist.
What is community cloud? A fifth model where several related orgs share a private cloud (e.g., all branches of a government). Rarely called out on 200-901; know it exists.
Is SaaS a deployment model? SaaS (Software as a Service) is a consumption model, not a deployment model. SaaS runs on top of public (usually) or private cloud.