EC2 Auto Scaling

You will need to use step scaling to have control over both the alarms and number of instances added/removed (the amount the desired capacity will be changed by the policy). Target tracking will dynamically decide how much capacity to add or remove based on the specific state of the group and metric at any given time; and does not allow you to customize the alarm settings. So while target tracking does work very well for the majority of use cases, if you have a hard requirement for your environment to only add/remove 1 instance at a time with a 5 minute alarm, then it won't work for you.

I highly recommend not setting both the scale-in and scale-out thresholds to 50% though, you should leave a gap between them to prevent oscillation. For example: lets say you have 3 instances, and the CPU goes to 60%, triggering the +1 step scaling policy. If the load stays constant, it will now be distributed to all 4 instances and the average CPU will drop to around 45% and the scale-in alarm will go off. This will then keep happening in a loop until the load goes up or down enough for one of the alarms to stay in the alarm state and the ASG reaches the min or max

Hi, you could use ASGAverageCPUUtilization to "Create a target tracking scaling policy (console)" as per https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-scaling-target-tracking.html#available-metrics You could create a scale-out and scale-in depending on the CPU utilisation