In March 2009 Amazon Web Services (AWS) announced an innovative new pricing model for buying cloud compute capacity, Reserved Instances (RIs), which enabled customers to make an upfront commitment at a resource level with two key benefits:
- Discounted pricing – cheaper hourly rates
- Guaranteed capacity – compute capacity available when needed
Over the years AWS has continued to iterate on the RI pricing model and in March 2017 AWS launched their latest improvement, Reserved Instance Size Flexibility. This new feature simplifies RI planning and improves the flexibility in how you can use your RI discounts. This automatically enables an RI to apply to any size of compute instance you run inside of a instance family type.
You don’t have to take any action in order to benefit from RI flexibility but it is important to understand how they work, and the impact for you as a user of Orbitera.
Understanding Reserved Instances
RIs are a discount, they are not usage of compute. Basically by making an upfront commitment you have bought a discount vs on-demand pricing. This is a big benefit when you have sustained workloads that are predictable and planned. When you purchase a traditional RI you specify the following parameters:
- Location (availability zone)
- Type (family)
- Size (e.g., Small, Large)
- OS (e.g., Linux, Windows)
When you run compute instances that match the parameters of your RI, the usage is automatically discounted at the hourly RI rate.
The major drawback of traditional RIs is that your usage has to exactly match the parameters of RI or you will not receive the discount. If your RI is in US West 1 and you launch an instance in US West 2 then you will be charged on-demand. This is challenging as you have to make an upfront commitment for 1- 3 years and lock into an RI, but your business will evolve and your needs will likely change over that time.
AWS made this a little simpler in September 2016 with the announcement of Regional RIs. Instead of specifying Location at an Availability Zone (AZ) (e.g., US West 2), you could now have the flexibility to specify a Region (e.g., US West). Now usage from any AZ within the Region that matches the other parameters automatically receives the RI discounts. However, AWS does not guarantee capacity for Regional RIs, so there is a potential risk.
How does RI Instance Size Flexibility work?
Flexible RIs make it a little easier to plan by removing the Size constraint on RIs. Now any size instance within an instance family (Type) can benefit from the RI discount. You don’t have to worry about what size instance you will need.
AWS is able to do this by normalizing all compute hours, the RI discount application is then based on the normalized units of usage. Regardless of instance family (type), the normalization factor is as follows:
|Instance Size||Normalization Factor||Instance Size||Normalization Factor|
Let’s say you already own an RI for a t2.xlarge. This RI now applies to any usage of a Linux/UNIX t2 instance with shared tenancy in the region.
This could be:
- One t2.xlarge instance,
- Two t2.large instances,
- Four t2.medium instances,
- Eight t2.small instances,
- Or half of a t2.2xlarge instance.
You read that right, an RI can apply to “half’ an instance because the usage is normalized. In the example, half of the t2.2xlarge will be billed at the RI hourly rate, and the other half will be billed at the on-demand rate.
This is a powerful feature for RI users! As an AWS user with a traditional t2.xlarge RI, you only realized the benefit when you ran a t2.xlarge instance. With Regional RI and Instance Size Flexibility you can run any combination of instances from the t2 instance family (Type) and still get your RI benefit.
Flexible RI’s make your capacity planning simpler, but there are important considerations to keep in mind:
- Not all compute instances are eligible for RI Flexibility
- Currently AWS only supports Regional RIs. That means if you bought your RI before Regional RI’s launched in September 2016 then they are not eligible for instance flexibility.
- Only instances running Linux/UNIX are supported. Windows OS is currently not supported.
- Your RI must have shared tenancy enabled.
- Is guaranteed capacity important to your business? If so, then you may not want RI Instance Size Flexibility. When you purchase Regional RI’s you give up guaranteed capacity which is only available with traditional AZ based RIs.
- Flexible RIs are flexible, but only within the same instance family (Type). You can’t apply your t2 RI to m3 usage. Therefore, you still need to plan what type of instance family matches your business needs.
- Flexible RIs automatically apply to any usage across multiple linked accounts via Consolidated Billing. This means that you don’t have to purchase the RIs in the account that is using the instance.
- If your business needs have changed, you can modify your existing RI purchases (i.e. Traditional RI can be converted to Regional RI). This doesn’t change the term of your RI, but modifies the parameters.
Whether you are an AWS Distributor, Reseller or end Customer, Orbitera will automatically apply the Reserved Instance Size Flexibility benefit based on your RI Inventory and usage.
The Orbitera platform normalizes your eligible usage and, based on the RI Inventory available to your account, will share the RI benefit to ensure the lowest possible bill.