SAP-C02 Practice Question 5: Multi Region Active Active Serverless
AWS Certified Solutions Architect – Professional (SAP-C02) · Test 1 · question 5 of 10
A gaming company is designing a new serverless leaderboard API built on Amazon API Gateway, AWS Lambda, and Amazon DynamoDB. Players are concentrated in North America and Europe, and the product team has set the following requirements: players must be served from the Region closest to them to minimize latency; writes must succeed in either Region, because a player's score can be posted from either continent; the service must keep accepting reads *and* writes if an entire Region becomes unavailable, with recovery measured in seconds and no human intervention; and the team does not want to build or operate any replication code.
Eventual consistency between Regions of a few seconds is acceptable, and the application already resolves last-writer-wins conflicts at the item level. Which architecture should the company choose?
Show answer and explanation
Answer:
DynamoDB global tables give a fully managed, multi-Region, multi-active table: every replica accepts reads and writes, changes propagate asynchronously (typically within a second), and conflicts are settled last-writer-wins. That satisfies "writes in either Region" and "no replication code to build". Pairing it with a Regional API Gateway plus Lambda stack in each Region and Route 53 latency-based routing sends each player to the closest healthy Region, and attaching health checks to those latency records makes Route 53 stop returning a Region that fails its health check, so failover is automatic and DNS-TTL fast.
Option A keeps a single active Region. AWS Global Accelerator improves the network path but does nothing for Regional failure when there is only one endpoint group with data, and restoring from hourly backups is an RPO of up to an hour and an RTO of many minutes, nowhere near "seconds, no human intervention".
Option B rebuilds global tables by hand. Bidirectional Lambda replication over DynamoDB Streams is the classic anti-pattern: it loops (a replicated write generates another stream record unless carefully tagged), it has no conflict resolution, and it is exactly the replication code the team said it would not operate.
Option D uses the right data layer but the wrong routing. A Route 53 failover record sends 100 percent of traffic to the primary Region, so European players pay a transatlantic round trip in normal operation, which violates the latency requirement. It also wastes the multi-active capability of the global table.
Reference: Global tables: multi-Region replication for DynamoDB