A company is launching a new application on an Amazon ECS cluster using the Fargate launch type. The company monitors CPU and memory because it expects high traffic at launch but wants to reduce costs when utilization decreases. What should a solutions architect recommend?

The correct answer is to use AWS Application Auto Scaling with target tracking policies. This service is specifically designed to automatically adjust the scalable capacity of various AWS resources, including Amazon ECS services on Fargate. Target tracking policies allow you to define a desired utilization level (e.g., 70% CPU utilization) and Application Auto Scaling will add or remove tasks to maintain that target, effectively managing costs during fluctuating traffic. Using Amazon EC2 Auto Scaling is incorrect because Fargate is a serverless compute engine, meaning you don't manage EC2 instances directly. Scaling at scheduled times might not be optimal for unpredictable traffic spikes. Using a Lambda function for scaling is an overly complex and less efficient solution compared to native Application Auto Scaling.