An international company will deploy VPCs in us-east-1 and eu-west-2 plus on-premises data centers in the US and UK. They plan two Direct Connect links (one in each country), transit gateways in each Region, Direct Connect gateways, and transit VIFs attaching the Direct Connect gateways to the transit gateways. Transit gateways will be peered. The design must make traffic follow the shortest geographic path so on-prem traffic uses the local Direct Connect when available, and must automatically fail over using the private WAN if a Direct Connect link fails (and vice versa). How should the engineer configure which prefixes are advertised on the transit VIFs to meet these requirements?

The correct approach is to advertise local VPC-specific CIDR prefixes and an aggregate route for the entire AWS environment. This allows traffic originating from the local data center to prefer the specific, more granular routes for local AWS resources, ensuring the shortest path. If the local Direct Connect fails, the data center will then use the aggregate route, directing traffic over the private WAN to the other Direct Connect link, which will then route it to the appropriate AWS Region. Advertising only the aggregate route would prevent granular routing and local path preference. Advertising all specific VPC CIDR blocks from both Regions would lead to suboptimal routing, as traffic might traverse the private WAN unnecessarily even when a local Direct Connect is available. Using BGP communities with regional aggregates is less efficient than advertising specific local routes for path preference and an aggregate for failover.