Reserved Instances

Reserved instances, often called RIs, are one of the oldest and most widely used ways to reduce cloud compute costs. A customer agrees to pay for a particular instance type, operating system and region for a one-year or three-year term. In return, the provider charges a much lower hourly rate than on-demand pricing. The discount applies automatically to matching usage, so engineers do not need to change how they launch resources. The term is most closely associated with Amazon Web Services, while Microsoft Azure offers a similar option called reservations, and Google Cloud offers committed use discounts.

The trade-off is commitment. The customer pays for the reservation for the full term whether or not it uses the capacity, so a reservation that no longer matches the organization's workloads becomes a cost rather than a saving. For IT financial management, reserved instances are a major lever for reducing compute costs, and managing them well requires good forecasting of steady usage, careful purchasing and regular monitoring.

Definition of Reserved Instances

A reserved instance is a billing commitment, not a physical server. It is a contract to pay a discounted rate for a defined amount of compute usage with specified attributes, such as instance family, size, operating system, tenancy and region, over a one-year or three-year term.

When running instances match those attributes, the reserved rate is applied instead of the on-demand pricing rate. Unmatched usage continues to be billed at on-demand rates.

Terms and Payment Options

On AWS, reserved instances are available for one-year and three-year terms. Three-year terms offer larger discounts, which AWS describes as up to 72% for standard reserved instances compared with on-demand prices, depending on the instance type and payment option.

Customers can pay all upfront, partially upfront or with no upfront payment. Paying more upfront generally increases the discount. Upfront payments are usually recorded as prepaid expenses and amortized over the term.

AWS offers standard reserved instances, which give the largest discounts but limited flexibility, and convertible reserved instances, which can be exchanged for different instance types in return for a smaller discount.

Scope and Flexibility

Regional reserved instances apply their discount to matching usage anywhere in a region, and for many Linux instance types they apply across sizes within the same family. This flexibility makes them easier to use fully.

Zonal reserved instances apply to a specific availability zone and also reserve capacity there, which can matter for organizations that need guaranteed capacity, but they are less flexible.

Standard reserved instances that are no longer needed can, in some cases, be sold to other customers on the AWS Reserved Instance Marketplace, which provides a way to recover part of their value.

Importance of Reserved Instances in IT Financial Management

Compute is usually the largest part of a cloud bill, and a large share of compute usage in most organizations runs continuously. Paying on-demand rates for this steady usage is one of the most common sources of avoidable cloud cost.

Reserved instances can reduce that cost substantially without any change to applications. Because the discount is a billing change, engineers can keep running workloads exactly as before.

They also shift some costs from variable to committed, which affects budgeting, forecasting and accounting.

Savings Potential

For steady workloads, such as production databases and core application servers, reserved instances can cut compute costs by a large fraction. Across an organization's eligible usage, the result can be millions of dollars a year.

Because savings come from billing rather than engineering changes, they can be achieved quickly, which makes reserved instances one of the first rate optimization steps in many FinOps practices.

Savings are measured by metrics such as commitment coverage and effective savings rate, which show how much eligible usage is covered and how much less the organization pays than on-demand rates.

Risks

The main risk is paying for reservations that go unused, for example when workloads move to different instance types, are migrated to containers or serverless services, or are retired. Low commitment utilization can erase the savings.

Standard reserved instances are tied to specific attributes, so architectural changes during the term can leave them stranded. Three-year terms increase this risk, since more can change over a longer period.

These risks have led many organizations to prefer more flexible options, such as savings plans, for much of their compute usage.