Commitment Discounts

Commitment discounts exchange purchasing flexibility for a lower unit price. Instead of paying only the standard on-demand rate for resources as they run, an organization accepts a defined financial obligation over a term. The commitment may cover particular resource characteristics or a specified amount of eligible spending per hour. The central decision is whether future usage is predictable enough to justify that obligation.

In the cloud, commitments generally operate as billing arrangements rather than instructions to deploy resources. Eligible usage receives the discounted rate according to the agreement's matching rules. A commitment does not necessarily create a virtual machine, keep an application running, or guarantee that capacity will be available. Resource deployment, billing discounts, and capacity assurance are separate concerns, although some purchasing options combine them.

AWS illustrates two common approaches. Amazon EC2 Reserved Instances provide billing benefits for matching EC2 usage, with applicability determined by attributes and offering rules. AWS Savings Plans instead require a consistent amount of eligible hourly spending for one or three years. Compute Savings Plans can apply across eligible EC2, AWS Fargate, and AWS Lambda usage, while EC2 Instance Savings Plans are tied to an instance family in a selected Region. Neither model means that every charge on the account receives a discount.

Consider an application that continuously needs ten virtual machines but scales to thirty during business peaks. An engineer might purchase commitments for a conservative portion of the ten-machine baseline, retain On-Demand capacity for uncertain growth, and use Spot capacity for interruptible background processing. Committing to the peak would leave part of the obligation unused during quieter hours. Rightsizing resources and removing idle workloads before purchasing commitments reduces this risk.

On-Demand purchasing is useful for short-lived workloads, experimentation, and demand that cannot yet be forecast. It avoids a long-term commitment but normally has a higher unit rate than comparable committed usage. Spot Instances use spare EC2 capacity at discounted prices and can be interrupted; their discount is not earned through a long-term commitment. They fit workloads such as checkpointed batch processing, not applications that cannot tolerate interruption without additional resilience.

Capacity Reservations address a different requirement: access to EC2 capacity in a particular Availability Zone. An On-Demand Capacity Reservation does not itself provide a commitment discount, although eligible Reserved Instance or Savings Plans benefits can apply to its charges. Regional EC2 Reserved Instances provide a billing discount without reserving capacity; zonal EC2 Reserved Instances can also provide a capacity reservation for matching attributes. Exam questions often test this distinction between reducing price and assuring capacity.

Dedicated Hosts and Dedicated Instances primarily address hardware isolation, compliance, or licensing requirements rather than commitment discounts. A Dedicated Host provides a physical server dedicated to a customer, with host-level visibility and placement control. Dedicated Instances run on hardware isolated from other customers but do not provide the same host-level control. Dedicated Host Reservations can reduce eligible host costs through a term commitment; choosing dedicated tenancy alone does not create a commitment discount.

Payment schedules and commitment obligations must also be distinguished. All-upfront, partial-upfront, and no-upfront options may be available depending on the offering, but no upfront payment does not mean no commitment. Stopping resources generally does not stop the financial obligation. Engineers should track utilization, meaning how much purchased commitment is consumed, and coverage, meaning how much eligible usage receives a discount. High coverage is not a success if excessive commitments remain unused.