Cloud Comparison Cloud Updates Migration

The Ultimate Guide To AWS Reserved Instances, Part 1: Planning And Analysis

Reserved Instances don’t always get the recognition they deserve as one of the best ways to reduce spend in AWS. This is due to the common misperception that taking advantage of RIs is an inherently complicated and time-consuming process. Unfortunately, this fear results in many organizations foregoing one of the essential “3 R’s” of lowering AWS costs. That’s a big mistake, since EC2 purchasing is often the largest monthly cost for AWS customers.

In reality, RIs don’t have to be complicated. By making a 1- or 3-year commitment to purchase AWS capacity, you can achieve significant savings over traditional, pay-as-you-go procurement. Better yet, CloudHealth Reserved Instances Lifecycle Management takes the hassle out of RI management by providing the modeling, modification, and optimization capabilities needed to feel confident about your purchase decisions.

Here’s some good news: after working with hundreds of customers and helping them purchase millions of dollars of RIs (CloudHealth customers purchase more than $100 million in EC2 RIs every month), we’ve developed a simplified approach to make the process less intimidating. Follow along with this blog series to learn the five phases of managing and optimizing Reserved Instances:

  • Phase 1: Reserved Instance Planning & Analysis
  • Phase 2: Rightsizing Your Infrastructure
  • Phase 3: Making the Reservation Purchase
  • Phase 4: Ongoing Reserved Instance Monitoring & Modification
  • Phase 5: Reserved Instance Chargeback

Read on for a deep dive on Phase 1: Reserved Instance Planning and Analysis. Or, you can jump ahead at any time and read our ebook The Ultimate Guide to Amazon EC2 Reserved Instances.

Phase 1: Reserved Instance Planning & Analysis

It’s easy get caught up in the complexity of planning your first RI purchase––the sheer number of options at your disposal can be overwhelming at first blush! When planning your first purchase, you should take care to group like instances and evaluate costs by group.

Step 1: Group Similar Amazon EC2 Instances

To effectively manage RI purchasing at scale, you should start by grouping instances by one or more criteria, such as environment, application type, operating system (OS), or instance family. One popular approach is to group instances by their functional purpose (Production, Development, etc). Or better yet, start by analyzing just one instance family, like all of your r4 instances or in one Region. Another approach is to consider analyzing by OS (Windows, Linux, etc), since AWS charges different costs and offers different features based on the OS. Typically I see customers going with their biggest (and most expensive) grouping first.


Fig. 1  |   Most of the On Demand Instance costs are in Production, so maybe I should start there with my RI purchase


Step 2: Evaluate Instance Groups

After breaking out your infrastructure into functional groups, you should determine the most cost-effective use of RIs by identifying which groups have the biggest impact on your total AWS costs. RIs are best suited for “always-on” infrastructure, so you may choose to exclude infrastructure groups that are on less than 65% of the time.


Fig. 2  |   These i3s scale up and down a lot throughout the day. If they are running on average less than 65% of the time, I might not want to buy RIs for them

Step 3: Determine Your Reserved Instance Purchase Strategy

What’s your plan for this upcoming RI buy? Will you be basing the purchase on a set budget (e.g., $50,000) or on a projected coverage rate (e.g., 70% of my total running instance hours in a month should be covered by reservations)? Both are valid approaches! You also need to decide how often you will be making your purchases, is it monthly, quarterly, or continuously? Best practice is to purchase continuously, as needed. Lastly, will you be making delegated or centralized purchases? Because Reservations will have affinity to the account in which they were purchased, it’s an important consideration. Best practice is typically to purchase from the consolidated master account.


Fig. 3  |   The CloudHealth optimizer allows you to plan your purchase either based on RI coverage or budget. You can also choose which accounts to purchase from, which groups to limit your analysis to, and many other variables

Step 4: Understand the Business Context of Your Reserved Instance Buy

Every Reservation actually has eight different features that require a decision point from you. To help you make good decisions around location, scope, class, type, OS, (for background on all these terms, check out this blog post: Convertible or Standard class RIs) consider the following questions, on a group by group basis:

  • What percentage of instances within this group do you reasonably expect will still be running 1 or 3 years from now?
  • Will the instances in this group remain within their current region?
  • Am I likely to switch the instance type family (e.g. m4 to C5) for instances in this group?
  • Am I likely to switch operating systems over the next 1 or 3 years?
  • Do we need capacity assurance for any of our instances?
  • How much capital can we commit to this buy right now? What is my budget?

Answering these questions will help you determine which RI type, scope, and class you should be purchasing for this group of instances.


Fig. 4 | In this example, I can see that historically, we’ve used a lot of r3 instances, but it looks like we are making the switch to r4s. I should likely not buy any RIs for the r3 family.

Next step, Rightsizing Your Infrastructure!

This critical but often overlooked step when making an RI purchase can have a large impact on your bottom line.

Part 2: Rightsizing Your Infrastructure!

To learn more about how you can cut down your AWS spend with Reserved Instances, check out The Ultimate Guide to Amazon EC2 Reserved Instances eBook!

Get Your Free Copy