Lowering Costs with Reserved Instances

AWS has multiple models for purchasing services on its platform. These different models aren't designed to be confusing, but at times can be. If you're not adjusting the pricing model of the services, you utilize you could very well be handing AWS free money at the end of every month.

We'll be covering EC2 (Elastic Compute Cloud, essentially Virtual Machines) Reserved Instances (RIs), what they are, when you should use them, and how much it could save you.

What are Reserved Instances?

AWS has three different primary pricing models for EC2, On-Demand, Reserved, and Spot - Reserved instances are for when you're aware that you'll be utilizing that instance type 24/7 for the next year or three. Your organization gives AWS a commitment, and in return, AWS lowers the machine's monthly cost.

This works because it allows AWS to better plan out capacity in data centers, allowing them to pass those cost savings along to you.

How do they work?

Over the years, this has changed, but there are two primary models with subclasses in each of these models.

The first model is "Standard" this allows you to purchase an RI knowing that it will be the exact capacity you require for the next 1 Year or 3 Years, and they cannot be modified.

An example of a use case for Standard is that you've had your company's website running on AWS for the last year, and you're aware that at all times you must have at least one m5.large instance type - your application leverages autoscaling, but you'll always need that base instance. You'd select a Standard 1 year RI for an m5.large, and it would automatically associate the reserved instance with the running instance and decrease the bill.

Inside of the standard model, there are three different sub-options, each with a different price based on your ability to give AWS a commitment.

  • Full Upfront - This is the highest cost savings available for RIs; you would immediately pay for the reserved instance upon requesting it for the next full year.
  • Partial Upfront - This will be a slightly lower cost-saving, but you'll only pay a small amount upfront and have a lowered monthly bill for the instance
  • No Upfront - This will result in the highest pricing but the lowest amount of upfront investment.

The second model, Convertible, gets a little more confusing than that - but it provides you even more flexibility for your environment. Initially, RIs could not be modified or changed, which resulted in either wasted spend because the instance type was no longer in use because of a requirement to swap AZs, networking types, or instance families. To keep the explanation short, Convertible RIs essentially give you "RI Credits" that can be used across instance families, sizes, and availability zones.

To visualize it, let's say that your organization rents a fleet of large delivery box trucks, about 10 of them, and you committed to a contract for two years of rentals. Over time you've noticed that your box trucks are half empty, and you're spending a lot of money on fuel without maximizing the utilization - you've decided you want to use smaller vans for deliveries. Instead of renting 10 box trucks, the rental company allows you to switch and rent 30 delivery vans instead of 10 box trucks. Your capacity remains the same; the cost would remain the same, but the design you've laid out the capacity changes.

Convertible RIs have the same pricing model but lower savings due to the added flexibility for the organization - this is great for an environment that's always in flux but wants to realize some of these savings. The key thing is that due to this, you'll need to keep a close eye to ensure that your RIs match your actual instance types and you're not leaving unused capacity on the table.

Why use them?

Reserved Instances allow you to lower your monthly cost on EC2 Instances up to 72%, resulting in significant savings. If you're already using the instances and can predict your minimum requirements, it's a great opportunity to save.

Using Reserved Instances is not an all or nothing thing. You can continue to use separate On-Demand instances, scale your machine clusters with Spot instances if you'd like. RIs allow you to plan a baseline and should be used as a tool in your belt to pay AWS as little as possible.

How can we help?

CloudAccountant.io's signature CARE review process guarantees that we'll find savings in your account and help you better understand cost models in AWS. Schedule a call now to see how much we can save your organization.

Tune in next week for a little more on pricing models and how we can help you learn to save money!