Spot Instances on AWS are a great way to massively reduce the compute costs for your IaaS environment (AWS boasts of up to a 90% reduction in cost) not bad! So what’s the catch? Well that depends on the type of workload you’re planning to deploy to spot instances.
When creating an On-Demand workload in AWS you specify the size and have full control of the availability of the VM – you can have it running 9-5 Monday to Friday and shut it down automatically outside these hours if it’s a test environment for example or have it running 24/7 if it’s a critical business service. Spot instances flips this on its head – AWS control when this VM is up and down, it does this by analysing demand within its data centres . When there’s plenty of capacity it will fire up your spot instance and when AWS need the capacity back they’ll give you a 2 minute warning before taking the capacity back. You have full control over the Start-Stop features, for example when AWS require the capacity back they will pause your instance so it’s able to resume right where it left off.
This is clearly not aimed at production or critical workloads but is perfect for test & development environments, stateless web servers or analytics applications. A great feature and well worth a look if you have services and applications that would suite this.