A media firm serves personalized thumbnails from S3 and viewing history in DynamoDB to US and EU users, facing high inter-Region transfer bills from full replication of all buckets and tables. Only curated assets and active profiles need global access. Which replication design balances global performance with transfer cost?

The balanced design succeeds because prefix-filtered S3 Cross-Region Replication limits transfer and storage charges to only curated thumbnails while preserving low-latency access in both US and EU, with scoped storage classes preventing over-retention of derivative objects, and DynamoDB global tables providing multi-active reads and writes for active profiles without custom sync code. Using regional DynamoDB tables fronted by DAX fails by concept because DAX is a regional read-through cache that reduces read latency but cannot offer cross-region active writes, conflict resolution, or locality for EU users updating viewing history. Syncing DynamoDB with Streams and Lambda fails by concept because it creates fragile one-way eventual replication requiring custom error handling and ordering logic, lacking the managed multi-active consistency, automatic failover, and low write latency that active profiles require. Expanding identical replication to an extra APAC region fails by concept because it restores the original cost problem by paying inter-region transfer and storage for users who do not exist in scope, violating the requirement that only US and EU need global access. For exams, remember that caching and stream processing improve performance or enable migration but are not substitutes when the stem explicitly demands curated-scope replication and active-active data access, so always match region count and filter scope to stated users. As a practical check, validate prefix filters in S3 replication rules, confirm global table replica status and replicated write capacity in both regions, and track S3 replication and DynamoDB inter-region transfer in Cost Explorer before and after rollout.