The computing power (compute) is the basis of the Cloud. The performance of AWS Cloud services is well established: Amazon provides high quality products that meet the needs of most businesses, regardless of industry or size.
As is often the case, what is of concern in Cloud services is the vagueness surrounding the billing model. Here is a complete article to help you better understand the pricing of compute AWS products. We will discuss the products one after the other to give you a 360° view of AWS's tariff specificities.
To allow you to navigate more simply, here is the summary of this article:
7 - Free tools from the compute AWS range
1 - Amazon EC2 Pricing
Amazon EC2 is one of the flagship products in the AWS offering. Amazon EC2 is a virtual machine service that enables businesses to benefit from secure and flexible computing power, all with the simplicity and immediacy of Cloud services.
Paradoxically, Amazon EC2 pricing is much less fluid than AWS promises. Let's try to get a clearer picture.
EC2 instances are available in 4 major AWS billing models:
Billing on demand
This is the B-A-BA of AWS pricing: you only pay for what you consume, in real time. Simple and effective. Well, in principle. It would be really simple if pricing was not so variable. The main variability criteria: the operating system, the technical configuration of the instance (number of vCPUs, GB of RAM, type of storage...) and the deployment region.
The (very) wide range of instances proposed allows to cover an almost infinite scope of application. You can opt for instances that are more computationally oriented, or more storage oriented, machine learning, etc...
On-demand billing is suitable for development teams looking for real flexibility, for applications generating traffic peaks that are difficult to anticipate or for workloads that cannot be interrupted by chance.
The reservation of instances
Reserving AWS EC2 instances is a well-known practice among AWS users. As we have already presented in this article, the reservation of instances allows you to anticipate your consumption of Cloud resources and to benefit from very interesting discounts on your billing. The commitment is for 1 or 3 years and you benefit from a rate of up to -75% off the initial price (in theory, of course).
Reserved EC2 instances are suitable for very linear applications and for users requiring true stability over the medium or long term. The benefits are essentially to project over time, to establish a more accurate budget and to reduce the impact of variability in expenses on your AWS bill.
Saving Plans
Like reserved instances, Saving Plans are an effective way to anticipate your Cloud budget and limit the uncertainty associated with too much variability in spending. In concrete terms, Saving Plans are similar to improved instance reservation. The most? Additional flexibility with respect to families of instances, size, SO or region. Concretely, if you are currently using Reserved Instances, we advise you to migrate gradually to Saving Plans.
Like reserved instances, Saving Plans allow you to reserve computing power in advance, for the same duration: 1 or 3 years. Discounts are equivalent and allow you to save up to 72% compared to on-demand instances.
If you would like to know more about how to introduce Saving Plans into your Cloud strategy, we have prepared some tips on this article.
Spot instances
Spot EC2 instances allow AWS users to take advantage of unused computing power in the AWS Cloud. It's a kind of recycling of the "compute" available. The advantage of Spot instances is the discount percentages that cannot be compared with other types of billing. Take advantage of discounts of up to 90% off on-demand rates.
Please note, however, that Spot instances are not suitable for all uses. They are reserved for applications with flexible start and end periods, for users with urgent requirements for a large amount of additional computing capacity, and for static and fault-tolerant applications. Example: Training a Machine Learning model is a good use case.
Dedicated bodies
Dedicated instances allow AWS users to take advantage of dedicated hardware dedicated to the performance of the application concerned. Concretely, it is a physical EC2 server reserved for your own use. Costs can be reduced by using your existing software licenses (Windows Server, SQL Server, SUSE Linux Enterprise Server). You can both take advantage of the dedicated instances in on-demand billing or via the reservation system.
2 - Amazon Lambda pricing
AWS Lambda is also one of the services most used by AWS Cloud consumer web developers. For those who are not familiar with this service, and to make it simple enough, AWS Lambda allows you to deploy code without worrying about server management (serverless model). The advantage of Lambda is therefore above all the flexibility and fluidity brought to the development teams.
From a pricing perspective, AWS Lambda is purely pay-per-use. You only pay according to the use you make of the service. Note: the AWS Free Tier offer allows you to benefit from 1 million free requests. This is enough to discover the service with confidence. Two elements are taken into account in AWS Lambda billing: the number of requests and the code execution time.
1 - The number of queries made on your functions
A query is an event notification or request call on the function.
The price per request varies by region, but there are few variations. All regions USA, Canada, EU, Asia-Pacific and South America are based on a similar price. Only the Asia-Pacific (Hong Kong) and Middle East (Bahrain) regions are based on higher pricing.
2 - Code execution time
The duration is calculated from the moment the execution starts until it
to the moment it ends, rounded to the nearest 100ms.
Time pricing varies depending on the amount of memory you allocate to your feature and the region in which you deploy your feature. Of course, you are free to define your needs and analyze the budget needed to deploy your AWS Lambda instance.
Finally, this pricing offers you the possibility to optimize your strategy according to your needs. We advise you to calculate the best need/price ratio by taking into account these two aspects: is it more interesting for you to group your functions by increasing the execution time, or conversely, to divide your requests as much as possible to allow you to reduce the impact of the duration on your invoicing.
The case of allocated simultaneity
AWS Lambda now offers an additional option: allocated simultaneity. Specifically, AWS offers the ability to reserve ready-to-launch runtime environments to optimize the response times of your cloud-hosted functions. Lambda functions require a certain warm-up time, which can sometimes cause latency problems for some users. We also mentioned the subject in a full article that you will find right here!
Obviously, it's not on the house. The allocated simultaneity is charged according to the reserved capacity and the duration for which you configure it. Logically, you also pay for the requests and duration, following the model presented to you above. On top of all this, the price of this allocated simultaneity varies according to the regions, with much more flexibility than for classical Lambda instances. It's hard to figure out, isn't it?
So if we recap, it adds an extra line to your bill. Interesting, if you really need it. Essentially for users who need optimal responsiveness for their user experience. For the others, there is no need to add blur to an already difficult to define service.
Data transfers
How often in the Cloud, data transfers are charged. This is no surprise for AWS Lambda, which takes over the basic AWS data transfer system: incoming data transfers are free of charge, as are data transfers between two services in the same region. Outbound data transfers are charged according to a rate defined by AWS and variable by region.
Note that data transfers between the following services
Amazon S3, Amazon SQS, Amazon DynamoDB, Amazon Kinesis , Amazon Glacier, Amazon ECR, Amazon SES
ECR, Amazon SES, Amazon SNS or Amazon Simple DB and Amazon Lambda functions,
all within the same AWS region, are free.
Finally, to finish convincing you of the complexity of Lambda billing, note that it is possible to face additional costs related to the use of other AWS services (e.g. need for AWS S3 storage). In this case, additional charges may appear on your invoice and it will be necessary to take them into account beforehand.
3 - Amazon ECS Pricing
The Amazon ECS (Elastic Container Service) allows the orchestration of containers. This service allows the execution, shutdown and simplified management of Docker Containers in a cluster. In particular, it is compatible with AWS Fargate, the serverless computing service. This avoids server provisioning and management and simplifies the work of developers.
Amazon ECS pricing is very specific and is based on two models: the Fargate launch model and the EC2 launch model. Each of these billing models actually depends on the associated service.
3.A - The Fargate Launch Model
With this Fargate release model, billing depends on the quantity of vCPU selected and the memory used by the application running in containers. The vCPU and memory are calculated as soon as the images from the container are retrieved, up to the end of the job (rounded to the nearest second). You don't have the possibility to define a precise quantity of vCPU or memory: Fargate makes its own calculation according to your request.
3.B - The Amazon EC2 Launch Type Model
With the Amazon EC2 launch model, you don't have to incur any additional costs. You simply pay for EC2 instances on the principle of AWS EC2 pricing, or EBS storage, in the classic way. You can therefore take advantage of the reservation of instances, spot instances, Saving Plans or dedicated instances specific to Amazon EC2.
It is also possible to deploy Amazon ECS in an Outpost. You will be pleased to note that the pricing is exactly the same for a launch via EC2. Relieved?
4 - Amazon ECR Pricing
Amazon ECR (Elastic Container Registry) is a service that allows developers to easily store, manage and deploy Docker container images to save time. Concretely, it is a useful solution for deploying Docker containers quickly and efficiently. Some kind of big library of containers.
In terms of pricing, Amazon ECR is rather advantageous because it is available without any upfront fee or commitment. You only pay for the amount of data stored and the volume of data transferred to the Internet.
Storage fees are $0.10 per GB per month. A fixed price, for once. Incoming data transfers are free of charge. Outgoing data transfers are charged with a price that varies by region. Note that the price simply varies in the Asia-Pacific, Middle East and South America regions. Prices in the USA, Europe and Canada are the same.
It is also important to note that Amazon ECR offers a free level to new customers. For one year, after you sign up, you can benefit from 500 MB of free storage per month. An offer that will help most of you see things more clearly.
5 - Amazon EKS Pricing
Amazon EKS (Elastic Kubernetes Service) is a product quite similar to Amazon ECS, dedicated to the orchestration and deployment of containers. It is suitable for sensitive and critical applications with high requirements for security, reliability and scalability.
Pricing is, for once, based on a fairly simple principle: you pay $0.010 per hour for each Amazon EKS cluster created. AWS EKS is executable via Amazon EC2 and AWS Fargate. Pricing therefore depends on the pricing for these two services :
- Amazon Fargate: you only pay for the vCPU and the memory used, from the moment you start uploading your Kubernetes container image.
- Amazon EC2: you pay for your instances according to the billing method selected (on-demand, reserved instances, Saving Plans, spot instances or dedicated instances), not forgetting the cost of storage.
6 - Amazon Lightsail Pricing
Amazon Lightsail is a simplified service that allows developers to focus on the code and forget about the bill. Nice promise, right?
Indeed, it's a fact: pricing is really simplified. You choose your Linux or Windows environment, then one of the configurations pre-formatted by Amazon. Each configuration costs you a fixed price, nothing could be simpler. We wish everything was so intuitive. Of course, Amazon Lightsail is a much less flexible service in terms of configuration and is therefore not suitable for all uses. Indeed, Amazon Lightsail is a very standardized offer.
Please note that there may still be a peculiarity: data transfers exceeding the rate indicated in your package will incur overrun charges.
7 - Free tools related to AWS computing power
In the "compute" range, Amazon Web Services offers several free services that work with the help of the reference services in the range, such as Amazon EC2, AWS Fargate or AWS Lambda for example.
AWS Batch
AWS Batch is a tool that allows AWS users to run hundreds of thousands of batch calculation tasks on AWS. It's a service that has the advantage of being fully managed and integrated into Amazon, so you don't have to go through third party software or open source solutions. AWS Batch does not include any additional fees: you only pay for the EC2 instances or Lambda functions that you use in your batch deployments.
AWS Elastic Beanstalk
AWS Elastic Beanstalk allows you to deploy and scale web applications or services developed on Java, .NET, PHP, Node.js, Python, Ruby, Go and Docker on Apache, Nginx, Passenger and IIS servers. This service offers you great speed of execution and true simplicity of deployment. As with AWS Batch, AWS Elastic Beanstalk is available at no extra charge.
AWS Image Builder
Finally, AWS Image Builder is a service that allows you to simplify the creation, maintenance, validation, sharing and deployment of Linux or Windows Server images for execution on Amazon EC2. Based on the same principle as the two above-mentioned services, AWS Image Builder does not incur any additional costs.
Conclusion
It's hard to figure out, isn't it? However, we hope this article has helped you better understand the pricing of the Compute range on AWS. As it is difficult to remember all this information in a single reading, we advise you to keep this reference in your favourites, so that you can consult it whenever you want to know more about the pricing of one of the products in the Compute AWS range. To summarise, here is a table of the different essential information to be retained for each service mentioned in the article:
If you wish to go further, we have produced several articles on the major AWS product families.
