During a game launch, a company ran 12 r6g.16xlarge (memory optimized) Amazon EC2 instances behind a Network Load Balancer. Monitoring showed actual usage at about one quarter of the CPU and memory expected. Demand has now dropped and become more variable. The company will replace the fixed fleet with an Auto Scaling group that scales based on CPU and memory to minimize cost. Which Auto Scaling configuration is the most cost-effective while meeting the requirements?

The most cost-effective solution uses r6g.4xlarge instances because the original workload was memory-optimized (r6g) and actual usage was one-quarter of the r6g.16xlarge instances. A .4xlarge instance is one-quarter the size of a .16xlarge instance, making it an appropriate right-sizing choice. Setting minimum capacity to 3 ensures high availability and handles baseline demand (12 / 4 = 3 instances). Desired capacity matching minimum capacity provides a stable starting point, and maximum capacity of 12 allows scaling up to the original peak if needed. Incorrect options: c6g.4xlarge instances are compute-optimized, not memory-optimized, which doesn't match the original workload. m6g.4xlarge instances are general-purpose, also not matching the memory-optimized workload. r6g.8xlarge instances are too large (half the original .16xlarge), leading to higher costs than necessary given the one-quarter usage. The lower max capacity also limits scaling.