Bitzu

SAA-C03 ยท Elastic compute performance

Auto Scaling policies and metrics

Target tracking, step, simple, scheduled and predictive scaling, choosing metrics such as CPU or SQS queue depth per instance, cooldowns and warm-up.

Key points

  • Target tracking (recommended): keeps a metric near a target value.
  • Step scaling: steps sized by alarm breach. Simple scaling: one change, then cooldown (300 s default); others: instance warmup.
  • Scheduled: known times; predictive: forecasts daily/weekly patterns, launches early.
  • Metric must fall as capacity grows: average CPU, requests per target, SQS backlog per instance (not queue length).
  • Several policies firing: largest wins.

Source: https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-scale-based-on-demand.html

Practice questions

Single choice

Which EC2 Auto Scaling policy type keeps a chosen metric, such as average CPU utilization, close to a value you set?

  1. Simple scaling
  2. Scheduled scaling
  3. Target tracking scaling
  4. Step scaling
Show answer

Answer: C. Target tracking scaling

Target tracking adjusts capacity to hold a metric near a target value and creates the CloudWatch alarms for you. Step and simple scaling react to alarms you define with fixed adjustments, and scheduled scaling acts at set times regardless of metrics.

Select all that apply

Which metrics work well with a target tracking policy because they fall as the Auto Scaling group grows?

  1. Average CPU utilization of the group
  2. ALB request count per target
  3. SQS backlog per InService instance
  4. Total number of messages in the SQS queue
  5. Total requests arriving at the load balancer
Show answer

Answer: A. Average CPU utilization of the group, B. ALB request count per target, C. SQS backlog per InService instance

Average CPU utilization, ALB request count per target and SQS backlog per instance all drop roughly in proportion when you add instances, which is what target tracking needs. The raw queue length does not change with group size, and the total request count to the load balancer is driven by clients, not by capacity.

True or false

The Auto Scaling group's default cooldown pauses target tracking and step scaling policies after every scale-out, just as it does for simple scaling.

Show answer

Answer: False

False. Cooldowns gate only simple scaling policies. Target tracking and step scaling can scale out immediately; they rely on instance warmup so that newly launched instances are not counted before they are ready.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More SAA-C03 topics in this module