The correct answer is: Update the CloudFormation templates to switch from M5 to R5 EC2 instances. Deploy the Amazon CloudWatch agent on the instances to collect custom latency metrics for future tuning and planning.

R5 instances are memory-optimized and well-suited for workloads that require high-performance in-memory processing. Since the application is stateful and memory-bound, increasing memory availability directly addresses the bottleneck. Using CloudFormation to implement the change ensures consistency and automation. Deploying the CloudWatch agent enables the collection of fine-grained, custom metrics (such as application latency), which supports better long-term capacity planning and performance optimization.

Why the other options are incorrect:

Switching to T3 instances and using the AWS Management Console is not a good choice for memory-bound, high-performance applications. T3 is a burstable instance type meant for lightweight workloads and is unlikely to improve performance in a sustained high-demand scenario. Additionally, using the console lacks repeatability and control compared to CloudFormation.

Adding EC2 instances to an Auto Scaling group and manually adjusting capacity helps with scaling, but does not address the fundamental problem of memory constraints. Also, manual intervention is less operationally efficient than an automated, metric-informed solution.

Switching to R5 instances and relying only on built-in EC2 memory metrics may help, but Amazon EC2 does not natively expose memory utilization metrics. Without installing the CloudWatch agent, memory metrics and application-specific performance indicators (like latency) will be missing, limiting visibility and hindering optimization.

Why this question is important: This scenario tests your ability to recognize when instance type changes, rather than horizontal scaling, are the correct solution. It also emphasizes the importance of automated infrastructure updates using CloudFormation and proper monitoring practices with CloudWatch.

Tip: When applications are stateful and memory-bound, think in terms of vertical scaling and memory-optimized instances. For observability, always consider using the CloudWatch agent for custom metrics like memory and latency.

Full AWS Practitioner Certification Question