GCP Professional Cloud Architect Practice Question 3: Hybrid Connectivity SLA Topology

The documented multi-region topology for a 99.99% Dedicated Interconnect SLA is precisely this: at least four connections spanning two metropolitan areas, using both edge availability domains within each metro, with a minimum of four VLAN attachments across two regions (at least two per region) and a Cloud Router in each region. Global dynamic routing is required so that each Cloud Router advertises and learns routes for the whole VPC network rather than only its own region. Dedicated Interconnect also satisfies the private-path and throughput requirements: 20 Gbps of sustained traffic is comfortably within four connections, and nothing crosses the public internet.

Option B fails on both dimensions of the redundancy design. Placing both connections in one metro and the same edge availability domain means a single maintenance event or facility failure takes down all connectivity, and a single Cloud Router is itself a regional failure domain. This is below even the 99.9% topology.

Option C is a single point of failure regardless of what the partner's backbone looks like. A Google Cloud SLA covers the resources you provision; one attachment on one Cloud Router has no redundancy on the Google side, and Partner Interconnect SLA tiers have their own redundancy requirements that a lone attachment does not meet.

Option D can reach a 99.99% SLA for the HA VPN gateway itself, but it violates two hard constraints: the traffic rides the public internet, and per-tunnel throughput caps mean two tunnels cannot carry a sustained 20 Gbps with predictable latency.

Reference: Establish 99.99% availability for Dedicated Interconnect