What is an Amazon EC2 Reserved Instance and how does it work?

An Amazon EC2 Reserved Instance is a billing discount model in Amazon Web Services that offers substantial cost savings of up to seventy two percent compared to standard On-Demand instance pricing in exchange for a one-year or three-year usage commitment. Rather than provisioning a distinct physical or virtual server, a Reserved Instance acts as an automated pricing coupon applied by the AWS billing system to matching running On-Demand instances in your account. By committing to a specific instance configuration and term length, organizations with steady-state or predictable workloads can significantly reduce their cloud computing expenses and optionally reserve compute capacity within a chosen Availability Zone.

Understanding the Billing Discount Mechanism

A common misconception is that purchasing an Amazon EC2 Reserved Instance creates a new virtual machine. In reality, Reserved Instances exist purely at the billing layer. When you launch an EC2 instance, you always launch an On-Demand instance. Every hour, the AWS billing system evaluates the attributes of your running instances and compares them against the active Reserved Instances in your account or AWS Organization.

When a running instance matches the attributes of a Reserved Instance, AWS automatically applies the discounted hourly rate instead of the standard On-Demand rate. Conversely, because you are committing to continuous utilization over the contract term, you are billed for the Reserved Instance whether or not you have a matching EC2 instance running.

The effective hourly cost of a Reserved Instance can be expressed using the following formula:

\[\text{Effective Hourly Rate} = \frac{\text{Upfront Payment}}{\text{Total Hours in Term}} + \text{Discounted Hourly Usage Fee}\]

The total percentage saved compared to running an equivalent On-Demand instance over the same duration is calculated as:

\[\text{Discount Percentage} = \left( 1 - \frac{\text{Total RI Cost}}{\text{On-Demand Hourly Rate} \times \text{Total Hours in Term}} \right) \times 100\%\]

Core Attributes of a Reserved Instance

For the billing discount to apply, the running EC2 instance must match four core attributes defined at the time of the Reserved Instance purchase:

Instance Type: The specific family and size of the instance, such as m5.large or c6i.xlarge.
Platform Description: The operating system and software package, such as Linux/UNIX, Red Hat Enterprise Linux, SUSE Linux, or Windows Server (with or without SQL Server).
Tenancy: Whether the instance runs on shared hardware (default tenancy) or single-tenant hardware (dedicated tenancy).
Region or Availability Zone (Scope): The geographical scope where the discount applies.

Regional vs. Zonal Scope and Capacity Reservations

When configuring a Reserved Instance, you must choose between a Regional scope and a Zonal scope, which changes both how the discount is applied and whether compute capacity is guaranteed:

Regional Reserved Instances: Apply the billing discount to matching usage across any Availability Zone within the specified AWS Region. Regional RIs do not provide a capacity reservation, meaning if an Availability Zone experiences capacity constraints, your ability to launch new instances is the same as standard On-Demand usage. However, Regional RIs provide instance size flexibility.
Zonal Reserved Instances: Locked to a single, specific Availability Zone. In addition to the billing discount, Zonal RIs provide an explicit capacity reservation. AWS guarantees that capacity will be available for you to launch the matching instance type in that specific Availability Zone whenever you need it. Zonal RIs do not offer instance size flexibility.

Instance Size Flexibility

Regional Reserved Instances running on the Linux/UNIX platform (with default tenancy) offer a feature called instance size flexibility. This allows the billing discount to automatically scale up or down to cover different sizes within the same instance family based on a normalization factor.

Every instance size has an assigned normalization factor. For example, a small instance has a factor of $1$, a medium has a factor of $2$, a large has a factor of $4$, and an xlarge has a factor of $8$. The total normalized units provided by a reservation are determined by:

\[\text{Total Normalized Units} = \text{Number of Instances} \times \text{Normalization Factor}\]

If you purchase one c5.xlarge Regional Reserved Instance (which equals $8$ normalized units), the billing engine can apply that single discount to any of the following running simultaneously in the same Region:

Two c5.large instances ($2 \times 4 = 8$ units)
Four c5.medium instances ($4 \times 2 = 8$ units)
One c5.large and two c5.medium instances ($4 + (2 \times 2) = 8$ units)
Half of the usage of a single c5.2xlarge instance ($16$ units required, so $50\%$ runs at the RI rate and $50\%$ is billed at the On-Demand rate)

Reserved Instance Offering Classes

AWS provides two primary offering classes for Reserved Instances, balancing discount depth against architectural agility:

Standard Reserved Instances

Standard RIs provide the highest possible discount (up to $72\%$ off On-Demand pricing). They are best suited for stable, steady-state workloads where the instance family and operating system will not change over the commitment period. While you cannot change the instance family or operating system of a Standard RI, you can modify the Availability Zone, scope (between Regional and Zonal), or instance size within the same family.

Convertible Reserved Instances

Convertible RIs provide a slightly lower discount (up to $66\%$ off On-Demand pricing) in exchange for the ability to exchange the Reserved Instance during the term for another Convertible RI with different attributes. You can change the instance family, instance type, platform operating system, and tenancy, provided the new reservation is of equal or greater total monetary value than the remaining value of the original commitment.

Term Commitments and Payment Options

The exact discount percentage you receive depends on two financial levers selected during purchase:

Term Length

One-Year Term: Commits for $31,536,000\text{ seconds}$ ($365\text{ days}$) and offers a moderate discount for medium-term visibility.
Three-Year Term: Commits for $94,608,000\text{ seconds}$ ($1,095\text{ days}$) and provides the deepest available discount for long-term foundational infrastructure.

Payment Options

All Upfront: You pay for the entire one-year or three-year commitment in a single payment at the time of purchase. Your recurring monthly hourly usage fee becomes $\$0.00$, yielding the maximum discount rate.
Partial Upfront: You pay a portion of the total commitment upfront and pay the remainder via a discounted hourly rate billed monthly over the term. This offers a mid-tier discount rate.
No Upfront: You pay $\$0.00$ at the time of purchase and instead commit to paying a discounted hourly rate every month for every hour in the term. While this yields the lowest discount of the three options, it preserves capital while still saving heavily over On-Demand pricing.

When using Consolidated Billing in AWS Organizations, Reserved Instance billing discounts are shared across all member accounts by default. The AWS billing system first applies the Reserved Instance discount to matching running instances within the specific account that purchased the reservation. If the purchasing account does not have enough matching usage in a given hour to fully consume the reservation, any unused discount capacity automatically floats to other accounts in the same organization that are running matching instances, maximizing utilization and cost efficiency across the entire enterprise.