Why you should use SPOT with AWS ASG

If you use AWS ASG (Auto Scaling Groups), you should use SPOT instances.

Cost Efficiency: Spot instances offer significant savings over On-Demand or Reserved Instances. On average, SPOT is about a third of the price of On-Demand and half the price of Reserved Instances.

Optimal Resource Utilization: AWS ASG is designed for dynamic workloads by adjusting the number of instances to meet the desired capacity. By incorporating SPOT Instances, you can tap into unused AWS capacity, making the most of available resources and minimizing infrastructure costs.

I hear you already, but what if you can not get a SPOT instance for a critical workload? You can ensure being able to get a SPOT instance by doing these two things:

- Select a range of instance classes that map to different physical server types. You can figure this out by reading the details of each instance class.

- Use as many availability zones (AZs) as possible.

If done correctly, I’ve never seen an AWS ASG SPOT configuration not be able to get an instance. The only edge case where this might not work are the big instance types like 10x and above and GPU. But for the paranoid, I have a solution to launch an on-demand instance when a SPOT launch fails.

Interested in what others think?