Simple Scaling Policies
What you'll learn: How to create straightforward scaling rules that add or remove instances when a single alarm triggers.
What is a Simple Scaling Policy?
A simple scaling policy performs a single scaling action when a CloudWatch alarm breaches its threshold. Think of it like a thermostat with just one setting: "If temperature exceeds 75°F, turn on the AC." There's no complexity—one trigger, one response.
Unlike step scaling (which has multiple thresholds) or target tracking (which continuously adjusts), simple scaling waits for an alarm to fire, makes its adjustment, then enters a cooldown period before it can act again.
How It Works
- You define a CloudWatch alarm based on a metric (e.g., CPU > 70%)
- The alarm breaches when the condition is met
- The policy executes once, adding or removing a fixed number of instances
- Cooldown period begins, preventing further actions until it expires
Example Scenario
Your Auto Scaling group has a minimum of 2 instances and maximum of 10. You create:
- Alarm: Average CPU utilization > 70% for 5 minutes
- Simple Scaling Policy: Add 2 instances when alarm triggers
- Cooldown: 300 seconds (5 minutes)
When CPU spikes to 75%, the alarm fires, two instances launch, and the policy waits 5 minutes before it can scale again—even if CPU remains high.
Key Limitation
The cooldown period can be too rigid. If demand keeps climbing, simple scaling won't respond until the cooldown expires. That's why step scaling (which skips cooldowns) is often preferred for dynamic workloads.
Key Takeaway: Simple scaling policies trigger one adjustment action per alarm breach, then wait out a cooldown period—making them best for predictable workloads with gradual changes.