Question 40
Domain 4: Accelerate Workload Migration and ModernizationA SaaS company runs a customer-facing application that requires a relational database with the following characteristics: sub-10ms read latency for a global user base, automatic failover with less than 30 seconds of downtime, support for up to 500,000 read requests per second, and the ability to scale read capacity across multiple AWS Regions. Cost optimization is a secondary concern. Which database solution meets these requirements?
Correct answer: B
Explanation
Amazon Aurora Global Database is built for global, low-latency reads, with “replication across multiple AWS Regions” and read scaling through Aurora replicas. It also provides fast regional failover, typically “less than 30 seconds,” which matches the downtime requirement while supporting very high read throughput for a customer-facing relational workload.
Why each option is right or wrong
A. Amazon RDS Multi-AZ with cross-Region read replicas in each target Region
RDS read replicas help scaling, but standard RDS is less optimized for global relational failover and extreme read scale.
B. Amazon Aurora Global Database with Aurora replicas in each Region
Amazon Aurora Global Database is the only option that natively combines cross-Region read scaling with a single writable primary and multiple Aurora replicas in secondary Regions, which is what the prompt’s global sub-10 ms read-latency requirement is testing. Under the Aurora Global Database design, failover to a secondary Region is typically under 30 seconds, and Aurora’s storage/replica architecture is the one intended for very high read throughput—up to hundreds of thousands of reads per second when scaled with replicas—without sacrificing relational consistency.
C. Amazon DynamoDB global tables with DynamoDB Accelerator (DAX)
DynamoDB global tables are NoSQL, not a relational database solution despite excellent global low-latency reads.
D. Amazon Aurora Multi-AZ with ElastiCache Redis in each Region for read caching
Redis caching improves reads, but caching does not provide a globally replicated relational database architecture.