Network Connectivity Center: Hubs, VPC Spokes and Mesh vs Star Topologies

GCP Study Hub
November 13, 2026

Network Connectivity Center lets you attach VPC networks as VPC spokes to a hub. Google manages the connectivity between the spokes, with no tunnels, gateways or routing appliances, so the design avoids per-tunnel bandwidth limits and connects many VPC networks at full bandwidth. Two settings decide which networks can reach which: the hub's preset topology, mesh or star, and the export filters configured on each spoke. Choosing those two correctly is most of what a Network Connectivity Center design involves, and both come up on the Professional Cloud Network Engineer exam.

Hubs and VPC spokes

A hub can connect a large number of VPC spokes. That scale is one reason to choose Network Connectivity Center over VPC Network Peering, whose limit per VPC network is far lower. Because Google manages the connectivity, there are no routing appliances to build, scale or keep running, and no tunnels whose throughput caps the design.

Routes still don't allow packets. Attaching networks to a hub gives them routes to each other, but firewall rules must also permit the traffic before anything flows.

Mesh and star topologies

A hub uses a preset topology that decides which spokes exchange routes.

Mesh. All spokes belong to a single default group, and each spoke exchanges its subnet routes with every other spoke on the hub. Mesh fits when every VPC network must reach every other one directly, as long as the ranges don't overlap or the overlapping ones are excluded from export.

Star. Spokes in the center group exchange routes with every other spoke, and spokes in the edge group exchange routes only with center group spokes. Star fits a classic hub and spoke design, such as a shared services VPC network in the center group and application networks in the edge group, so the edge networks reach shared services but never each other.

Group membership is what decides reachability. A center group spoke reaches every other spoke, so the center group is for networks that every spoke is allowed to reach. An edge group spoke reaches only the center, so the edge group is for networks that must stay apart from each other. Export filters work within the topology rather than around it. They decide which of a spoke's ranges are shared, not which spokes exchange routes, so they can narrow what a topology shares but never widen it.

Export filters on VPC spokes

Export filters control which subnet ranges a VPC spoke shares with the hub. They are configured on each spoke, not on the hub.

  • Include export ranges list exactly which ranges a spoke shares. Listing only approved subnets keeps every other subnet out of the other networks' routing tables, which enforces a policy at the routing layer. Firewall rules alone can't do that, because they can block packets while every subnet route stays visible.
  • Exclude export ranges keep matching subnet ranges from being exported. One exclude range can cover several subnets, as long as each subnet falls inside it, so a single broad range can hold back a whole group of subnets while every other subnet is still exchanged.

Route control is one of the clearest differences from VPC Network Peering. Peering exchanges all subnet routes between the two networks, and it always exchanges subnet routes that use private RFC 1918 addresses, so its export controls can't hold back a range like that. A Network Connectivity Center spoke can.

Overlapping ranges and spoke creation

Network Connectivity Center rejects a new VPC spoke if it would export subnet ranges that overlap ranges already exported by other spokes. Because export filters live on the spoke, an overlap is handled when the spoke is created. With gcloud, the spoke creation command accepts --exclude-export-ranges with the overlapping CIDR, so that range is never exported. This works as long as the excluded subnet doesn't need to be reachable from other spokes.

Hybrid spokes and a routing VPC network

Hybrid connections can join the hub too. VLAN attachments can be linked to a hub as a hybrid spoke. A dedicated routing VPC network that holds the Interconnect VLAN attachments and Cloud Router lets the hub learn on-premises routes dynamically and share them with every spoke, so on-premises connectivity is built once and reused by every attached network.

When to choose Network Connectivity Center

For the Professional Cloud Network Engineer exam, Network Connectivity Center fits when many VPC networks need to be connected without appliances or tunnel limits, when each network's visible routes must be controlled at the routing layer, or when a hub and spoke design must keep spokes isolated from one another. The details that decide the configuration are the topology, mesh for any-to-any and star for hub and spoke, and the export filters on each spoke, include ranges to share only approved subnets and exclude ranges to hold back overlapping subnets or any that shouldn't be shared. Firewall rules are still required on top of both.

Our Professional Cloud Network Engineer practice exams cover Network Connectivity Center alongside VPC Network Peering and hybrid connectivity, with detailed explanations that drill these distinctions.

Get tips and updates from GCP Study Hub

arrow