Three Key Steps to Making the Most of AWS Reserved Instances

AWS Reserved Instances can be powerful, cost-saving tools for your business. But before you start working with Reserved Instances (or RIs), it’s important to understand how they work, and the many factors that must be considered before you buy. Use these best practices to improve cloud efficiency for your enterprise.

AWS Reserved Instances can be powerful, cost-saving tools for your business. But before you start working with Reserved Instances (or RIs), it’s important to understand how they work, and the many factors that must be considered before you buy.

Recapping the risks and rewards associated with RIs

Before we move forward, let’s start with a quick RI primer, so we’re all on the same page. We like to think of RIs as really complicated coupons for cheaper hourly AWS rates on EC2, RDS, Redshift and ElastiCache. When you reserve instance hours, like with EC2 for example, you can run those instances at a discounted hourly rate.

However, there’s a bit of strategy and a commitment you need to make to lock in these cheaper hourly rates. The strategic part of an RI requires you to reserve hours within a specific AWS Region and Availability Zone (AZ) (e.g., US-West-1). The commitment is that you need to reserve every single hour within either a one-year or three-year term (depending on the RI, e.g. Convertible RIs are only available in three-year terms). Those reservations are also only applicable to the specific kind of instance (e.g., one year of discounted t2.micro hours are only applicable to t2.micro instances in the us-west-1 Region and AZ).

Getting the most savings out of RIs requires adhering to a continuous process of monitoring your AWS costs and usage to ensure you’re using as many RI hours as possible (since you pay for all of those hours ahead of time) as well as staying mindful of when it’s time to modify reservations (in the case for users running Linux on EC2) or purchase new ones. RIs can definitely save you money, but only if your infrastructure is using as many of those reserved hours as possible.

Comparing your RI options

You can gain even more savings by paying upfront fees to lock in lower rates. However, this means a higher cost ahead of time for long-term savings. This can get complicated, so understanding how much of your compute you actually use can really help determine if RIs are a good option, and which types you should buy. Using a cloud cost management tool, like Cloudability, makes this process much smoother.

Cldy emerge 2017 01 ec2 comparison matrix
Here’s an example of Cloudability’s EC2 Reserved Instance Comparison Matrix. It maps out RI terms and their potential savings rates so you can understand what you’re getting into before committing to an RI purchase.

Let’s walk through a few key steps throughout the term of an RI that are critical to understand in order to save as much as possible on AWS.

Step 1: Making the right RI purchases

Purchasing RIs is a great savings option for businesses who have enough operational history and data to know that their infrastructure will use as many of those reserved hours as possible. So if this sounds like your AWS infrastructure, great! You may proceed.

For many businesses, especially those starting out, a key question is, “Are you ready to purchase RIs?”

RIs may not be the best cost optimization tool for organizations that have no insight into what their cloud computing workloads consistently look like. In the case of EC2, imagine a team starting with a C4 instance, buying an RI for that C4 instance, then discovering only a month later that an M4 instance was a better fit. They can migrate their services to the M4, and stop paying for the C4 instance, but they still need to do something about the remaining reserved hours from the original C4 RI purchase.

Another example could be a team that only does a few hours of compute a week. It’s likely cheaper to run the instances that have a light workload at the On-Demand rate, rather than purchasing a year or three years’ worth of hours just for a cheaper rate.

Cldy emerge 2017 01 ec2 on demand reserved

This kind of high-level look at EC2 breaks down usage by On-demand and Reserved hours. While most costs in this example are using reserved hours, there’s always room for improvement to save more (e.g., hours that are billed On-Demand).

There will always be a mix of services that run on reserved or on-demand hours, depending on the kind of infrastructure you manage and services that you provide. There’s no right or wrong answer since every business is different. But in the game of savings, you want to have as many qualified services on reserved hours as possible to spend efficiently on AWS.

So if you have not first dug in and fully understood your business’ computing usage, that has to be the first step before buying RIs. For more insight on this, check out our comprehensive e-book on buying and managing RIs.

What about Convertible RIs and Regional Scope for EC2?

Convertible RIs are an offering class for EC2 introduced in 2016 that balances flexibility with savings to optimize your costs as your compute environment changes over time. They are only available in three-year terms, which means making a large commitment to EC2 usage. However, this type of RI allows you to exchange the RI for another to help you conform to shifts in your AWS EC2 infrastructure. With this ability to exchange also come fees for the exchange and less savings over time compared to other options, but lower hourly rates compared to On-Demand alone.

Regional Scope is a new option for EC2 RIs that launched along with the Convertible RI type. Previously, RI purchases required choosing a specific AZ (as described above), which had the benefit of reserving the hour of service in that AZ (also known as reserving capacity).

Selecting Regional Scope for an RI will allow it to cover any matching instance within the Region instead. While Regional Scoping grants flexibility in allowing your reserved hours to serve any matching instance within the Region, you give up reserving specific capacity in an AZ. One of our Technical Account Managers explored reserving for capacity and the kinds of tradeoffs involved in a previous RI piece.

Even with the prospect of flexibility, understanding the operational history and expectation of how much you’ll use the computing resource is still key to maximize Convertible RIs or Regional Scoping. In short, do the usage homework first!

Step 2: Using as many reserved hours as possible

Let’s say you’ve purchased a few RIs for your infrastructure: are you keeping tabs on how well you’re utilizing these hours? If your engineering team needs to switch their EC2 instances to a different size or family to accommodate the extra load from the growing popularity of your app or service, those original RI hours might not be applicable anymore (since they’re for a completely different instance family or size).

You need to keep a constant eye on both your AWS usage hours and your RI portfolio, even down to the individual RI, to ensure that your resources, such as EC2 instances, are using as many reserved hours as possible.

Ultimately, if no one in your company is going to utilize an RI, you should consider modifying it (for EC2 users running Linux) to fit smaller instances, exchanging unused EC2 RIs for usable ones if they’re of the Convertible type, or selling unused RIs on the AWS Marketplace. For a thorough look at RI planning, modifications, selling RIs and maintaining a healthy RI portfolio, we strongly recommend that you read our e-book on RIs.

Step 3: Keeping an eye on RI expiration dates

Keep a watchful eye on those RI expiration dates. Waking up to a cost spike in your next AWS bill could be a very nasty surprise.

Cldy emerge 2016 12 ri expiration 2

Using a cloud cost management tool to track RI coverage and term expiration dates can help you maintain a healthy RI portfolio.

As RIs get close to their expiration dates, this is the time to get engineering and operations together to talk about how to optimize RI coverage. Going over how many reserved instance hours were actually used throughout the terms and how the infrastructure has changed can provide some critical insights as to how your RI renewal and purchasing strategy might change.

This may seem like a ton of trouble for one measly RI, but the reality is that most companies using AWS at scale have hundreds to thousands of them across various AWS resources. This is when having a cloud cost management tool starts to become mission-critical.

Onward towards big savings!

Wrangling an RI portfolio is a big job that can yield quite a large payoff of savings if done correctly. One way to make this process easier is to try a cloud cost management solution that can help you make sense of your entire RI portfolio. If you want to see this in action with your own data, sign up for a free trial and get started saving today! RI portfolio management is an ongoing and ever-important task, and Cloudability helps simplify it by giving you a detailed view into how well your company uses your RIs.

目次

カテゴリ

タグ