Hub-Spoke vs Virtual WAN¶
Azure network topology decisions often come down to a self-managed hub-spoke model or Azure Virtual WAN as a managed connectivity fabric. Both patterns centralize connectivity, but they differ in scale model, control responsibilities, and operational burden.
Hub-spoke topology¶
In hub-spoke, shared connectivity and control services live in a hub virtual network while application workloads sit in spoke virtual networks.
Typical hub services include:
- Azure Firewall or NVA
- VPN Gateway or ExpressRoute gateway
- Shared DNS or inspection services
- Bastion and management services
Strengths¶
- Clear separation between shared network controls and application VNets
- Flexible for many landing zone designs
- Familiar to many Azure teams
Challenges¶
- Route management and peering complexity grow with scale
- Shared hub components can become bottlenecks or change hotspots
- Operational ownership remains largely with the customer
Azure Virtual WAN¶
Azure Virtual WAN provides a Microsoft-managed transit architecture for branch, remote user, and VNet connectivity.
Strengths¶
- Simplifies large-scale connectivity scenarios
- Reduces some routing and transit management burden
- Useful for distributed branch and global network scenarios
Challenges¶
- Less design freedom than fully self-managed topology
- Cost model may not fit small estates
- Teams still need clear policy, segmentation, and route intent
Topology comparison¶
flowchart TD
A[Connectivity need] --> B{Scale and complexity}
B -->|Moderate, customized| C[Hub-spoke]
B -->|Large, distributed, managed transit| D[Azure Virtual WAN]
C --> E[Hub VNet with shared services]
C --> F[Spoke VNets]
D --> G[Virtual WAN hub]
D --> H[Branches, VNets, users] Decision criteria¶
| Criterion | Hub-spoke signal | Virtual WAN signal |
|---|---|---|
| Estate size | Small to large, but manageable custom topology | Large-scale distributed connectivity |
| Custom routing and inspection | High need | More standardized managed model |
| Team network maturity | Strong customer-managed networking | Preference for managed transit |
| Cost sensitivity | Can optimize for smaller footprints | Better justified at larger scale |
Connectivity patterns¶
Hub-spoke fits well when¶
- Application landing zones need clear segmentation.
- Shared services are limited and highly customized.
- The organization already operates Azure Firewall, custom DNS, or NVA patterns centrally.
Virtual WAN fits well when¶
- There are many branches, remote users, or globally distributed network edges.
- Transit management overhead is becoming the problem.
- Standardized connectivity is more valuable than bespoke topology tuning.
Common anti-patterns¶
- Building a complex hub-spoke mesh when a simpler managed transit model would reduce risk.
- Choosing Virtual WAN without understanding route intent and segmentation needs.
- Centralizing all traffic inspection in a single bottleneck hub with no scale model.
- Over-peering environments without clear ownership or route documentation.
Evidence and trade-offs¶
- [Documented] Hub-spoke is a standard Azure reference topology for centralized connectivity.
- [Observed] Operational complexity rises quickly when peering, route tables, firewalls, and DNS all evolve independently.
- [Inferred] Cost should include shared appliances, gateways, cross-region traffic, and support burden.
- [Validated] Connectivity failover and route-change drills should be tested before broad rollout.
When not to use hub-spoke¶
- The organization mainly needs managed large-scale transit rather than custom central network services.
- The team cannot sustain route, firewall, and DNS operations centrally.
When not to use Virtual WAN¶
- The environment is small and cost-sensitive.
- Highly customized transit and inspection behaviors are mandatory.
Microsoft Learn reference¶
- https://learn.microsoft.com/en-us/azure/architecture/networking/architecture/hub-spoke
Takeaway¶
Use hub-spoke when you need customer-controlled shared network services and explicit landing-zone segmentation. Use Virtual WAN when connectivity scale and transit operations are the real problem to solve.