Study ANS-C01 multi-VPC and multi-Region connectivity for Transit Gateway, VPC peering, PrivateLink, VPC sharing, Cloud WAN, segmentation, and IP overlap.
Multi-VPC and multi-Region questions on ANS-C01 are about routing domains, not just connectivity. The exam expects you to decide whether the design needs point-to-point VPC peering, centralized Transit Gateway routing, narrow PrivateLink service exposure, shared subnets through VPC sharing, or global policy through AWS Cloud WAN.
The current AWS task statement is to design a routing strategy and connectivity architecture that includes multiple AWS accounts, Regions, and VPCs to support different connectivity patterns. The official scope names VPC peering, Transit Gateway, AWS PrivateLink, VPC sharing, IP subnets, and solutions for overlapping IP addresses. That means the strongest answer usually controls both the path and the blast radius.
| Design question | Why it matters |
|---|---|
| Is this two VPCs or a fleet of VPCs? | Point-to-point peering may work for two VPCs but becomes hard to operate at scale. |
| Does traffic need transitive routing? | VPC peering is not transitive; Transit Gateway and Cloud WAN are built for hub-style routing. |
| Is the requirement broad network routing or one private service? | PrivateLink can avoid full network-level trust. |
| Do accounts need to share subnets rather than connect separate VPCs? | VPC sharing supports centralized network ownership with application-account deployment. |
| Are CIDR ranges overlapping? | Normal routing cannot distinguish overlapping destinations without NAT, redesign, or service-level exposure. |
| Is segmentation required? | Route tables, attachments, segments, and endpoint policies must prevent unintended lateral movement. |
| Is the pattern regional or global? | Transit Gateway is Regional; inter-Region peering or Cloud WAN may be needed for global architecture. |
If an answer only says “connect the VPCs,” it is probably not ANS-C01-grade. You need to know who can reach what, through which route table, in which Region, and with what failure behavior.
| Requirement | Strongest first fit | Why |
|---|---|---|
| Two or a few VPCs need simple direct private connectivity and no transitive routing | VPC peering | Point-to-point, straightforward, and low operational overhead for small topologies. |
| Many VPCs, accounts, or hybrid paths need centralized routing | Transit Gateway | Hub-and-spoke routing, attachments, route tables, propagation, and segmentation. |
| One service should be privately consumed without broad VPC-to-VPC routing | AWS PrivateLink | Provider-consumer model exposes a service without opening the whole network. |
| Multiple accounts should deploy resources into centrally owned subnets | VPC sharing | Network account owns VPC/subnets; participant accounts place resources in shared subnets. |
| Multi-Region routing needs central global policy and segmentation | AWS Cloud WAN | Core network and segments are stronger when global policy is central to the requirement. |
| Existing SD-WAN needs AWS integration | Transit Gateway Connect or Cloud WAN integration | Overlay and appliance integration may matter more than normal VPC peering. |
| Overlapping CIDRs prevent direct routing | PrivateLink, NAT, address redesign, or carefully isolated route domains | Direct routing cannot resolve ambiguous destinations. |
Keep this model in mind when a scenario includes many accounts, VPCs, and Regions.
flowchart LR
A["Account / VPC boundary"] --> C["Connectivity pattern"]
C --> R["Route domain: peering, TGW table, Cloud WAN segment, or endpoint policy"]
R --> S["Segmentation and inspection"]
S --> E["Evidence: flow logs, Network Manager, route analysis"]
The exam often hides the real issue in the route domain. Two architectures can both connect the same VPCs, but only one may preserve production/nonproduction separation, inspection, or least-privilege service exposure.
| Exam clue | VPC peering | Transit Gateway | PrivateLink |
|---|---|---|---|
| Connectivity shape | One-to-one VPC relationship | Hub-and-spoke network routing | Producer-consumer service access |
| Transitive routing | No | Yes, through route tables and attachments | Not broad routing; service endpoint model |
| Scale | Good for small, simple topologies | Better for many VPCs, accounts, and hybrid attachments | Better for many consumers of one service |
| Segmentation | Per-peering route table and security design | TGW route tables, associations, propagation, inspection patterns | Endpoint service permissions and endpoint policies |
| IP overlap tolerance | Weak | Weak for direct routing | Stronger when only service exposure is needed |
| Common wrong use | Building a peering mesh for many VPCs | Adding a hub when two simple VPCs are enough | Using it when full bidirectional network routing is required |
This comparison is central to ANS-C01. The correct answer is not always Transit Gateway. It is the service whose connectivity model matches the requirement.
VPC sharing lets a central networking account share subnets with participant accounts in the same AWS Organization. Application teams can create supported resources in shared subnets while the networking team keeps control of VPC-level constructs such as CIDR ranges, subnets, route tables, network ACLs, and gateways.
Use VPC sharing when:
Do not use VPC sharing as a generic replacement for Transit Gateway, PrivateLink, or peering. It solves a different multi-account problem.
Transit Gateway is Regional. Multi-Region designs need an explicit cross-Region pattern.
| Requirement | Strong design cue |
|---|---|
| A few Regions need connected TGW routing | Transit Gateway peering with explicit route tables. |
| Global segmentation and policy are central | AWS Cloud WAN core network and segments. |
| A service should be consumed privately across boundaries | PrivateLink or service-level exposure pattern, depending on Region support and architecture. |
| Edge entry is global but backend routing stays regional | CloudFront, Route 53, or Global Accelerator with regional VPC design. |
| Disaster recovery needs controlled failover | Route 53 or Global Accelerator plus tested regional routing and data dependencies. |
Do not answer a multi-Region route problem with a single-Region Transit Gateway unless the design explicitly includes inter-Region peering or a global network layer.
ANS-C01 frequently makes “connect everything” the tempting wrong answer. In real architectures, the goal is usually controlled reachability.
| Requirement | Stronger pattern |
|---|---|
| Production and nonproduction must not route to each other | Separate TGW route tables, Cloud WAN segments, separate peering, or separate VPCs with explicit controls. |
| Shared services VPC must be reachable from many spokes | TGW hub route table or Cloud WAN segment with controlled propagation. |
| Internet egress must be inspected centrally | Central egress or inspection VPC with route symmetry and appliance scaling. |
| East-west traffic must pass through firewalls | GWLB or Network Firewall insertion with symmetric routing. |
| Only one service should be exposed to consumers | PrivateLink instead of broad routing. |
| Teams need common subnets but account-level ownership | VPC sharing. |
Route table design is part of security design. If the answer connects every attachment to one flat route table, question whether the scenario requires segmentation.
Overlapping CIDRs are an exam-grade constraint because normal IP routing cannot know which identical destination range is intended.
| Constraint | Better options |
|---|---|
| New environment, overlap can be avoided | Redesign CIDR plan before connecting networks. |
| Existing overlap, only one service must be reached | Use PrivateLink or an application/service proxy pattern. |
| Existing overlap, broader connectivity is required | Use NAT or renumbering where supported by the architecture and operational plan. |
| Overlap across acquired companies or partners | Keep route domains isolated and expose only required services. |
| Overlap inside a TGW or peering design | Do not assume propagation can fix ambiguous routes. |
If the stem mentions overlapping CIDRs, eliminate answers that simply add more routes without changing the addressing or exposure model.
| Scenario | Strong first design |
|---|---|
| Three VPCs in one account need simple full connectivity | VPC peering may be enough if there is no transitive requirement. |
| Hundreds of VPCs across accounts need centralized hybrid connectivity | Transit Gateway with route tables, RAM sharing, and controlled propagation. |
| Application teams need to deploy into centrally owned network subnets | VPC sharing. |
| SaaS-style internal service should be consumed by many VPCs without full routing | PrivateLink endpoint service and interface endpoints. |
| Global enterprise network needs policy-based segmentation across Regions | Cloud WAN or Transit Gateway inter-Region design, depending on scope. |
| Inspection must be centralized across many spokes | TGW route tables plus GWLB or Network Firewall inspection path. |
| CIDRs overlap between a partner and internal VPC | PrivateLink, NAT, proxy, or address redesign, not direct routing. |
| Stem clue | Eliminate first | Keep in play |
|---|---|---|
| “transitive routing” | VPC peering | Transit Gateway, Cloud WAN |
| “one service only” | broad VPC connectivity by reflex | PrivateLink |
| “shared subnets across accounts” | VPC peering | VPC sharing |
| “hundreds of VPCs” | full-mesh peering | Transit Gateway or Cloud WAN |
| “overlapping CIDRs” | more route propagation | PrivateLink, NAT, proxy, or redesign |
| “multi-Region routing policy” | one Regional TGW only | TGW peering or Cloud WAN |
| “segmented prod and nonprod” | one flat hub route table | separate TGW route tables or Cloud WAN segments |
When a multi-VPC or multi-Region path fails, use this order:
This order prevents the common mistake of debugging security controls while the selected connectivity pattern cannot support the requested path.
A company has 120 VPCs across several AWS accounts. Production and nonproduction networks must stay segmented, but both need access to a shared services VPC and to on-premises networks through Direct Connect. The team wants centralized route control and wants to avoid full-mesh peering. Which design is strongest?
Answer: Use Transit Gateway with segmented route tables.
Why: The scenario needs many VPCs, many accounts, centralized hybrid connectivity, segmentation, and avoidance of full-mesh peering. VPC peering does not support transitive routing and becomes operationally weak at this scale. VPC sharing is for shared subnet ownership, not routing between existing VPCs. Public load balancers violate the private controlled-routing goal.
Choose VPC peering for simple direct VPC-to-VPC connectivity. Choose Transit Gateway for centralized routing across many VPCs, accounts, and hybrid attachments. Choose PrivateLink for narrow private service exposure. Choose VPC sharing when accounts need to deploy into centrally managed subnets. Choose Cloud WAN when global policy and segmentation are the main requirement. Treat overlapping CIDRs as an architecture constraint, not a routing tweak.
Continue with Network Implementation to move from design patterns into building hybrid, multi-account, DNS, and automation paths.