AWS Certified Cloud Practitioner / ELB & ASG - Elastic Load Balancing & Auto Scaling Groups

Auto Scaling Groups (ASG) Strategies

๐Ÿง  AWS Auto Scaling Strategies

Every app has different traffic patterns , so you need the right scaling strategy to handle spikes, control costs, and keep performance steady โšก.

The basic flow: Traffic changes โ†’ Auto Scaling Group reacts โ†’ EC2 instances adjust automatically.


1. Manual Scaling

You change the desired capacity yourself (e.g., 1 โ†’ 2 instances).

  • Use for: testing ๐Ÿงช, small environments

  • Limitation: โŒ no automation โ€” someone has to act every time

2. Simple / Step Scaling โšก

CloudWatch alarms ๐Ÿšจ trigger a fixed action.

  • Example: CPU > 70% โ†’ add 2 instances โž•โž•; CPU < 30% โ†’ remove 1 โž–

  • You define the condition and the response

  • Best for: clear, well-understood thresholds

3. Target Tracking Scaling ๐ŸŽฏ (most important)

You set a target value for a metric, and the ASG keeps it there automatically โ€” adding instances when the metric rises, removing them when it falls.

  • Example: target CPU = 40%

  • Think of it like a thermostat ๐ŸŒก๏ธ: it just holds the temperature steady

  • ๐Ÿ”ฅ Exam tip: this is the easiest and most commonly used policy

4. Scheduled Scaling ๐Ÿ“…

Scale up or down at known times.

  • Example: scale to 10 instances every Friday at 5 PM

  • Use for: predictable spikes โ€” sales events ๐Ÿ›’, sports events โšฝ, business hours ๐Ÿข

5. Predictive Scaling ๐Ÿค– (advanced)

Uses machine learning to forecast demand from historical traffic ๐Ÿ“ˆ and scales ahead of the spike ๐Ÿ”ฎ, not after it.

  • Example: daily 6 PM spike โ†’ instances added before 6 PM

  • ๐Ÿ”ฅ Exam tip: keyword is "machine learning + forecasting"


โš–๏ธ Comparison

Strategy

Type

Trigger

Best Use

๐Ÿ–๏ธ Manual

Manual

User action

Testing

โšก Simple/Step

Dynamic

CloudWatch alarm ๐Ÿšจ

Threshold-based

๐ŸŽฏ Target Tracking

Dynamic

Metric target

Most common

๐Ÿ“… Scheduled

Time-based

Known schedule

Predictable traffic

๐Ÿค– Predictive

ML-based

Forecasting

Advanced automation


๐Ÿง  Quick recall

  • ๐Ÿ–๏ธ Manual โ†’ you're in control

  • ๐Ÿšจ Step โ†’ if X, then Y

  • ๐ŸŒก๏ธ Target Tracking โ†’ keeps things steady

  • ๐Ÿ“… Scheduled โ†’ time-based

  • ๐Ÿ”ฎ Predictive โ†’ forecasts the future