Besides picking instance types that provide (and cost) more than what you need, another common over-provisioning mistake is running too many instances in clusters or behind load balancers. Forgetting the on-demand aspect of the cloud becomes common – you don’t need to kick-off all the nodes of your cluster you may need on peak loads. You can add nodes as needed, and you can automate that too – AWS even has an auto-scaling functionality built in its platform.

Keywords: aws instance types, aws auti scaling, cloud computing