Customers who have Reserved Instance contracts with Amazon Web Services will be able to subdivide some of their Linux and Unix virtual machine instances while maintaining their capacity discounts, thanks to pricing changes announced Monday.
Reserved Instances allow customers to lock themselves into paying AWS for a certain amount of compute capacity with the company’s Elastic Compute Cloud (EC2) in exchange for a discount off its list price.
Under the new scheme, a customer could pay for a c4.8xlarge Regional Reserved Instance, and use that payment to run one c4.4xlarge and two c4.2xlarge virtual machines in one region instead of one massive VM. That’s a change from the past, when Amazon’s standard Reserved Instances locked customers into a particular virtual machine type.
In addition, customers that under-provision their Regional RIs can maintain their reserved instance discounts as they scale up. That means customers with c4.4xlarge reservations will end up only paying half of the on-demand price on top of the reserved instance price if they scale it up to a c4.8xlarge instance.
Customers don’t have to do anything to apply the pricing benefits to their instances. If they purchase a c4.8xlarge Regional RI and spin up two c4.4xlarge virtual machines in that region, the reservation will apply automatically.
That’s designed to be particularly useful for customers dealing with workloads that see spikes in demand (like retailers on Black Friday). They still get the reserved instance discounts but can maintain the scale-up flexibility that on-demand customers get.
It’s a move that may make Reserved Instances more appealing, because customers will get additional flexibility for the same amount of money. Customers who currently have Regional RIs with shared tenancy will see the benefit applied automatically.
It’s a move that builds on the RI changes AWS introduced last year. At that time, the company introduced Convertible Reserved Instances, which let customers take less of a discount over a three-year period in exchange for being able to switch instance types if they need to.
There are a couple major differences between the subdivision and scale-up capabilities introduced Monday and Convertible RIs. Subdivision flexibility only applies to instance size, not type. Meanwhile, Convertible RIs can be converted to other instance families if a customer chooses.
AWS’s latest changes also help deal with one of the key drawbacks Convertible RIs. In the past, customers who wanted to convert had to pick out instances of equal or greater value. Now, they can also choose to subdivide those instances.
Right now, this benefit is only available for Linux/Unix Regional RIs, so customers running other software like Microsoft’s Windows Server are still locked in. Customers who want to reserve dedicated capacity in a particular AWS Availability Zone won’t be able to take advantage of it, either.
The move comes less than a week after Google announced its own competing version of capacity reservation in what seemed like a response to AWS’s Reserved Instances. GCP Committed Use Discounts allow customers to get a discount if they commit to paying for a certain amount of CPU cores and RAM over the course of a one- or three-year contract.
Google’s plan offers more flexibility than AWS’s, because customers aren’t restricted to a set menu of different instance shapes. However, AWS’s cloud platform is the more popular one with customers at the moment.
This story has been updated to add a new fifth paragraph, specifying that Amazon customers don’t have to take action themselves to apply the new pricing benefits to their instances.