How does AWS auto-scaling work?

AWS EC2 Auto-scaling provides multiple instances of a selected EC2 class. The minimum and maximum number of instances is set by the user. Instances are added or removed according to a set of rules that is also user defined. For example, an instance might be added when CPU utilizations hit 80%, and instances removed when CPU drops below 60%. Much more sophisticated rules can be defined and crafted to best represent the particular workload.