AWS Certified Solutions Architect Professional SAP-C02 Practice Question
Customers in Europe report that the 95th-percentile response time for a latency-sensitive API sometimes doubles-from 180 ms to 350 ms-when traffic peaks. The service is hosted in us-east-1 on Amazon ECS tasks behind an Application Load Balancer (ALB), and clients currently resolve the production hostname directly to the ALB. The architecture team believes that routing traffic through AWS Global Accelerator (GA) could shorten network paths, but they want production data before deciding to adopt GA.
Business requirements for the test:
- Use real-user traffic and measurable metrics.
- Expose no more than 10 % of production traffic to the potential fix during the test period.
- Allow an immediate rollback with no application changes.
Which approach meets all of these requirements and provides the most reliable evidence of any latency improvement?
Enable HTTP/2 on the ALB and track the ALB's Latency and 5xxError CloudWatch metrics for two weeks. Keep the setting only if average latency improves.
Create a standard AWS Global Accelerator that uses the existing ALB as its endpoint. Add a weighted Amazon Route 53 alias record for the production DNS name that routes 10 % of queries to the accelerator's DNS name and 90 % directly to the ALB. Monitor CloudWatch latency metrics for both paths and adjust the weight only if results improve.
Increase the CPU allocation of the existing ECS service by 20 % using an AWS CLI --force-new-deployment and compare pre- and post-change Latency metrics.
Provision a new Network Load Balancer in Europe, place it behind a Global Accelerator, and immediately update the application's DNS record to the accelerator FQDN; delete the accelerator to roll back if latency does not improve.