An Auto Scaling group must add capacity quickly when CPU passes 70 percent, but must remove capacity slowly to avoid flapping. How should scaling be configured?
Show answer and explanation
Answer: B
Two independent step scaling policies let you define an aggressive response for the scale-out alarm, such as adding several instances for a large breach, and a conservative response for the scale-in alarm. Each policy carries its own cooldown, so new capacity can be added again quickly while removals wait long enough to prove the load has really dropped.
A single shared cooldown forces the same delay in both directions, which produces either slow scale-out or rapid flapping.
Why the other options are wrong
- A. With one cooldown value, a short setting lets scale-in remove capacity too eagerly and a long setting throttles the fast scale-out the workload needs.
- C. Scheduled scaling follows a calendar and cannot react to a CPU spike that happens outside the schedule.
- D. A simple scaling policy with a long cooldown delays every later scale-out action, so the group reacts too slowly under load.