marrakeshh - Fotolia


Dissecting AWS EC2 Reserved Instances for savings

It's important to understand the differences in AWS Reserved Instance tiers to get the most return on your investment.

AWS EC2 instances come in three pricing options: On Demand, Spot Instances and Reserved Instances (RI). Reserved Instances allow you to optimize AWS compute capacity at a discounted rate. But to reap the full cost benefits of RIs, AWS users need to adhere to their specific frameworks.

To get the most return on investment with AWS Reserved Instances, IT teams must select parameters such as instance type (r3.large), platform (Linux/UNIX, Windows or Red Hat Enterprise Linux), location (availability zone), tenancy (dedicated or default), and term (one year or three year). However, if you change RI configuration parameters, other than term, you won't benefit from the cost savings.

Why the old Reserved Instance model didn't work

The previous AWS RI pricing model left something to be desired. The massive upfront costs drove users away. Even if customers liked the RI framework, its CAPEX-heavy purchasing model outweighed why many turned to AWS in the first place -- OPEX optimization.

Previous AWS RI pricing levels were based on usage patterns: Light, Medium and Heavy Utilization. Light Utilization RIs were geared toward two- to five-month periods of AWS use; Medium RIs worked for five- to 10-month usage timeframes; Heavy RIs were created for continuous usage. This concept gave AWS customers the ability to save 20% to 60%, but came with a challenge. To see the cost savings, IT teams needed to properly match user experience with buying behavior. And if they didn't, they paid more. For example, if a user purchased a Light Utilization RI and ran it for 11 months -- longer than its recommended usage -- it would become more expensive than a Medium or Heavy Utilization RI.

Understanding AWS's current Reserved Instances model

In December 2014, Amazon announced three new RI pricing models: All Upfront, Partial Upfront and No Upfront. These models are similar to the Heavy Utilization Instances.

Cost-saving tips for Reserved Instances

  • Modify Reserved Instances (RI), change the region or instance type, or sell the instance.
  • RIs have capacity reservations. If you purchase an RI for any instance type, you can launch it in a specified zone anytime.
  • Sell unused Reserved Instances in the Marketplace. 
  • Take advantage of volume discounts when the total list value is more than $50,000.
  • For micro-instances purchased as RIs within the free tier of a new account, AWS will charges according to RI pricing.

And these new models bring AWS cost optimization to light. RIs are used to reward long-term customers for their commitment. Therefore, if you are only interested in using AWS instances for a few months, the on-demand pricing model is your best economical bet.

Here is a breakdown of the current AWS RI pricing models:

  • All Upfront: Users pay for the entire Reserved Instance term (one or three years) by making a single upfront payment. This achieves a very cost-effective hourly rate compared to On-Demand.
  • Partial Upfront: Users pay for a certain amount of a Reserved Instance upfront while the remainder is charged over the course of the Reserved Instance term (one or three years). This option does not require users to pay one lump sum upfront, and still provides a discounted RI rate.
  • No Upfront: This is for users who do not want to make a large, upfront capital investment and would prefer to pay for operational expenses (OPEX). With the No Upfront model, users don’t pay anything upfront, but instead agree to pay a discounted rate over the course of the one-year term. This option saves about 30% on costs compared to On-Demand pricing.

It's important to remember that -- for all three pricing options -- AWS charges users whether an instance is running or not. Figure 1 shows how each RI model compares to On-Demand pricing using c4.large instances.

Price comparison of Reserved Instances versus On-Demand instances using c4.large instances.

As the figure shows, if you only use the instance for five to six months (out of an entire year), On-Demand pricing is the best option. In the past, Light and Medium Utilization models would have been an ideal fit, but the current pricing model is more cost-effective for longer periods of use.

Because the cost for the Partial Upfront for three years is only a bit more than the same model for one year, it's more economical to purchase the three-year plan, even if you only plan to run the instance for a little over a year. If you know you will run the instance for two years, you'll want to go with the All Upfront three-year option.

About the author:
Ofir Nachmani is a business technology advisor, blogger and lecturer. Ofir's extensive experience in the world of business technology has made his critically acclaimed blog,, the go-to guide for modern technology startups and developers in the world of cloud computing. Today he advises organizations, leading them through new IT market modifications, while building and executing a modern go-to-market strategy.

Next Steps

Learn why AWS users want more flexibility with Reserved Instances.

Dig Deeper on AWS pricing, cost and ROI