Question 32
Domain 3: Continuous Improvement for Existing SolutionsA company is running a web application with On-Demand Amazon EC2 instances in Auto Scaling groups that scale dynamically based on custom metrics After extensive testing, the company determines that the m5.2xlarge instance size is optimal for the workload Application data is stored in db.r4.4xlarge Amazon RDS instances that are confirmed to be optimal. The traffic to the web application spikes randomly during the day. What other cost-optimization methods should the company implement to further reduce costs without impacting the reliability of the application?
Correct answer: D
Explanation
Reserved Instances reduce cost when the instance type and size are already known and steady, matching the “m5.2xlarge instance size is optimal” and “db.r4.4xlarge Amazon RDS instances are confirmed to be optimal.” Keeping only the minimum always-on EC2 capacity and reserving the RDS database lowers spend while Auto Scaling still handles random traffic spikes, preserving reliability.
Why each option is right or wrong
A. Double the instance count in the Auto Scaling groups and reduce the instance size to m5.large
Changing to smaller instances alters the tested optimal compute profile and may hurt performance or scaling efficiency.
B. Reduce the RDS instance size to db.r4.xlarge and add five equivalent^ sized read replicas to provide reliability.
Downsizing the primary database changes a confirmed optimal tier; read replicas do not replace primary write capacity.
C. Reserve capacity for all EC2 instances and leverage Spot Instance pricing for the RDS database.
Spot pricing is for interruptible compute, not a fit for a production relational database baseline.
D. Reserve capacity for the RDS database and the minimum number of EC2 instances that are constantly running.
Reserved Instances are the right fit when the workload has already been validated to use a specific EC2 size and RDS class, because the discount applies only if the family/size and Region are committed for a 1-year or 3-year term under the EC2 and RDS RI pricing models. Since the application still needs a baseline of always-on EC2 capacity and the db.r4.4xlarge database is steady, reserving only that minimum steady-state capacity captures the discount without changing the Auto Scaling behavior that absorbs the random daytime spikes.