Cloud Architecture
Auto Scaling Groups (ASG)
Understanding the mechanics of Target Tracking vs. Step Scaling policies to optimize performance and cost.
The Ingredients: Policy Inputs
Metrics
CloudWatch data points like CPU Utilization, Request Count per Target, or Network In/Out.
CloudWatch Alarms
The “Trigger” mechanism that alerts the ASG when specific thresholds are breached over time.
Cooldown Period
A rest period that prevents the ASG from launching or terminating additional instances before previous changes take effect.
Capacity Limits
Hard boundaries defined by Min, Max, and Desired capacity settings within the ASG.
The Scaling Process
Recommended
Target Tracking Policy
“Set it and forget it. Like a thermostat.”
- Simplest to configure
- Automatically creates CloudWatch Alarms
- Best for metrics that increase/decrease linearly with load
Advanced Control
Step Scaling Policy
“Custom responses for different levels of urgency.”
- Allows for “Aggressive Scaling” during spikes
- Multiple steps for different breach magnitudes
- No cooldown period needed (uses warm-up instead)