Skip to main content

AWS Auto Scaling

  • What it is
    • Automatically adds and removes capacity to match demand — the mechanism that delivers Elasticity.
  • Two things share the name
    • EC2 Auto Scaling — manages a group (ASG) of EC2 instances with a minimum, desired, and maximum count. Also replaces unhealthy instances automatically.
    • AWS Auto Scaling — a unified plane that scales EC2, ECS, DynamoDB, Aurora replicas, and Spot Fleets from one place, based on a target utilization.
  • Scaling policy types
    • Target tracking — “keep average CPU at 50%”. The simplest and most common.
    • Step scaling — add N instances per alarm threshold breached.
    • Scheduled scaling — scale up every weekday at 8am for known patterns.
    • Predictive scaling — machine learning forecasts load and scales ahead of it.
  • Two directions
    • Horizontal scaling (scaling out) — more instances. What Auto Scaling does.
    • Vertical scaling (scaling up) — a bigger instance. Requires a stop/start, not automated by ASGs.
  • Pairs with
    • load balancers — the ASG registers new instances with the load balancer so traffic reaches them.

Linked from