AWS Instance Purchase Models Explained

Amazon Elastic Cloud, Amazon EC2, is a widely used AWS service for compute capacity in the cloud. It offers the flexibility of instance type, quantity and purchase model choice. The flexibility allows for quick scaling of capacity as determined by user demand. The most widely used purchase option being On-demand Instance. Besides On-Demand two other Instance purchase models exist.


Amazon EC2 provides a broad range of Instance types for utilisation for differing use criteria. Through varying the attributes of the Instances according to required function, flexibility is enhanced. The Instances all differ in configuration of CPU, memory and storage and networking capacity, ensuring each are well aligned to its specific purpose when used correctly.

Instances are purchased for your application requirements and three purchase models are available to choose from, available for all instance types. These include On-Demand, Reserved and Spot Instances. As with the types of Instances providing the flexibility in use the purchase models provide flexibility when it comes to cost. Each option has its place and when used in the correct manner savings can be made.

It can become challenging to distinguish between all the Instance types as well as the EC2 Instance purchase models and the effects each may have on your finances. The assortment available ensures the necessary flexibility and scalability, which is welcomed. We would like to ensure the correct Instance use whereby the Instance chosen matches the purpose best and the most appropriate purchase model or combination of models so that we are attaining the most from our Instance choice at the best possible cost all of the time.

The challenge is to understand the three models costs and associated benefits well enough to know when to switch between them. The purchase model you choose will depend on your budget as well as your technical requirements, one model will differ and have advantageous over another.

The three purchase models Explained

On-Demand Instance Model

Reserved Instance Model

Spot Instance Model

Description Fixed price for a specific Instance type

Charged at an hourly rate until such time the Instance is terminated


Reserving EC2 Instances through upfront one-time payment to secure a lower server usage charge when those Instances are used Spot Instances provide the ability to acquire unutilised EC2 compute capacity without upfront commitment and at hourly rates typically lower than that of On-Demand
Features Allows for the increase or decrease of compute capacity when demand changes

Removes convolutions of planning, purchasing, and maintaining hardware

The flexibility of paying only for what you use


Function same as On- Demand but at reduced cost

When the Instance is purchased it is done specific to instant type and availability zone

If needs change you can request to move your Instance to another zone, or change to another Instance type.

You may also be able to sell your reserved Instance


To maximise usage rates of Amazon cloud resources, Spot Instances are created and Spot prices decided by Amazon

When the Spot price falls below your max bid, your Instance will launch and will continue to run until your bid is reached not exceeding it

When the Spot price exceeds your max bid, Instance will be terminated

You are unable to control the initiation and termination of a Spot Instance

The Spot Instance can start and end at any given time

When to use You require the low cost and flexibility of Amazon EC2 without any up-front payment or long-term commitment

For Applications that can not handle interruption or with short term, spiky, or unpredictable workloads

For Applications being developed or tested on Amazon EC2 for the initial time


You are able to plan ahead and make upfront payments thus further reduce compute costs

Applications tend to have steady and predicable usage requirements

Your Applications require reserved capacity


Initiation and termination times for Applications can be flexible

For running non critical applications

When only exceptionally low compute prices will make the Application viable

You require urgent computing needs for large amounts of added capacity

Available to all instance types Yes Yes Yes
Launch   priority Priority over Spot Instances but not over Reserved

In the unlikely event during periods of very high demand, there is potential for On-Demand Instances in specific   Availability Zones not to be available for launch until demand balances out.

Highest Priority

Launch Priority over On-Demand and Spot Instances

Your Instance will always be available for launch and within the availability zone of purchase

Lowest Priority

If Instance capacity becomes completely utilised Spot Instances can be terminated to provision On-demand or Reserve Instance launch requests

Pricing Fixed price

Hourly rate charge until Instance is terminated

Only pay for the Instances you use

Upfront one-time fee payable (options may be available) for one and three year terms

Reduced fixed usage price

Three payment options are available when you purchase a Reserved Instance, they are:

  • All Upfront option (pay entirety upfront thus largest discount received)
  • Partial Upfront option (low partial upfront payment followed by a discounted hourly rate for the instance usage until termination)
  • No Upfront Option (no upfront payment required followed by a lesser discounted hourly rate for the instance usage until termination)

The achievable savings depend on the payment option your choose

No upfront fee payable

Hourly rate is variable, you bid at a price you specify

Amazon EC2 sets a Spot rate for each instance type in each availability zone for a given period

The spot price will fluctuate based on supply and demand but you will never pay more than you specify in with your bid

You are billed the running variable spot price

Can be 75% lower than On-demand instance prices

Server uptime predictable Yes Yes No


The three models ensure all user requirements are accessible. The flexibility of On-demand Instances with no upfront costs or long-term commitments may be the right purchase model for some. However the other options available can offer a different form of flexibility and enable you to acquire immense quantities of compute capacity at a more affordable cost making any application development feasible.

It may be suitable to secure more than one purchase model dependant on your compute capacity and requirements to achieve the combined benefits.

It’s recommended that a staple set of On-Demand Instances be launched to safeguard the level of compute resources necessary for those applications that do not allow for unpredictable workloads and fluctuations in service, and to complement these with Spot Instances when the prospect occurs. Using a combination of the two purchase models offers the advantages from both approaches.

Reserved Instances allow for substantial savings up to three-quarters discount is attainable compared to On-Demand Instance costs. Moreover they provide assurance in that reserved Instances will always have priority over the others and thus will always be available and ready for launch when needed.

Like different Instance types are better suited to certain application function in comparison to others as are the particular purchase models better suited for certain tasks than others. To maximise the value from your purchase models take care when deciding on the approach to follow.

Leave a Comment

Your email address will not be published.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Scroll to Top