Which AWS service handles sudden EC2 demand spikes at lowest cost?

Why the Answer Is Correct

AWS Auto Scaling is the only option that actually changes capacity in response to demand: it monitors metrics such as CPU utilization or request count that you attach to a scaling plan, then launches or terminates EC2 instances to match the load. Because capacity is added during spikes and removed once the spike passes, the company pays only for what it needs, which is exactly the combination of responsiveness and lowest cost the stem requires. This elasticity is a core AWS Cloud design principle and applies to EC2 Auto Scaling groups across multiple Availability Zones. As Kaushal1987 explained, Auto Scaling "automatically adjusts the number of Amazon EC2 instances based on application demand," adding instances when demand increases and removing them when it decreases. ShaiTay offered the same reasoning in one line: it "automatically adjusts number of EC2 instances due to demand."

Why the Other Options Are Wrong

AWS Compute Optimizer is an advisory service: it analyzes utilization metrics and produces rightsizing recommendations for EC2 instances, EBS volumes, and Lambda functions, but a human or automation pipeline must act on those recommendations — it does not scale anything in real time. AWS Cost Explorer is a billing and cost-visibility tool for exploring spend by service, tag, or time period; it reports cost after the fact and cannot add capacity during a traffic spike. The AWS Well-Architected Framework is a set of architectural best practices and review questions, not a runtime service, so it can guide a design towards elasticity but cannot implement it. All three are useful complements to Auto Scaling, but none satisfies the requirement to respond to demand changes automatically.

Community Comment Notes

Both commenters converge on AWS Auto Scaling and frame it in the same way: capacity is added on the upswing and removed on the downswing so that cost tracks actual usage. Kaushal1987 additionally stresses that this "helps maintain performance while minimizing costs by using only the resources needed," which mirrors the two-part requirement in the question stem. ShaiTay's shorter confirmation repeats the demand-driven adjustment of instance count. Notably, neither commenter was distracted by the cost-related wording in the question, which is the most likely source of error for other candidates.