A company hosts a website with a backend service deployed in a primary AWS Region and a disaster recovery (DR) Region. A single Amazon CloudFront distribution fronts the site. The company created a Route 53 failover record with health checks pointing to the primary Region’s backend and configured that record as an origin for the CloudFront distribution. A secondary failover Route 53 record points to the DR Region backend. Both Route 53 records have a TTL of 60 seconds, but failover currently takes longer than one minute. Which design provides the fastest failover time?

Creating a CloudFront origin group with both regional backends and configuring origin failover within CloudFront provides the fastest failover. CloudFront's origin failover is designed for rapid switching between origins based on health checks, typically within seconds, without relying on DNS propagation. The existing setup's Route 53 failover is for the distribution's origin, not for the backend itself, and DNS propagation (even with a low TTL) introduces delays. Deploying a second CloudFront distribution adds complexity and still relies on Route 53 failover between distributions. Changing TTL to 4 seconds for Route 53 records is too aggressive and can cause issues, and DNS propagation will still introduce delays. Latency routing is for directing traffic to the lowest latency endpoint, not for failover.