Cloud Pricing Comparison: AWS vs Azure vs Google Cloud

Image Source

AWS vs Azure vs Google: An Overview

Amazon Web Services (AWS) is the world’s leading cloud computing platform. It provides Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS) offerings. AWS services can provide organizations with on-demand computing power, storage, application services, and content delivery services.

Microsoft Azure is Microsoft's public cloud computing platform. It provides cloud services such as computing, analytics, storage, and networking. Users can choose from these services to develop and extend new applications or run existing applications on the public cloud. Azure has an extensive PaaS offering and robust security features, integrated with Microsoft’s security products like Azure Active Directory and Azure Defender.

Google Cloud Platform is a collection of public cloud computing services provided by Google. The platform includes a variety of managed services for developing compute, storage, and applications that run on Google hardware. Google offers a simpler pricing model than the other providers and lower pricing in many service categories. In addition, it offers unique compute offerings, including the leading managed Kubernetes service and Tensorflow Processing Units (TPUs) for AI workloads.

Google Cloud, Azure, and AWS offer hundreds of different products. Each has its own service structure, technologies, and pricing models. You will likely have thousands of possible deployment combinations in each cloud. This can be a bit overwhelming and costs for each of these providers can be difficult to manage.

Thankfully, each of these providers has a pricing calculator and additional tools that can help you estimate and forecast costs. It is critical to use these tools prior to your migration, and on an ongoing basis, to ensure you keep costs in control.

AWS vs Azure vs Google Pricing

Pricing is one of the most important factors when choosing a cloud platform provider. Because all three providers have different pricing models and discounts, it is difficult to make a definitive comparison. Here is a brief summary of the pricing model for each provider:

• AWS:  Amazon has very complex pricing for some of its models, and provides dedicated tools like AWS Calculator, AWS Cost Explorer, and Trusted Advisor that can help you estimate costs and get opportunities for cost savings. At the same time, Amazon provides deep discounts on its services with multiple savings models, including spot instances, reserved instances, and savings plans.

• Microsoft Azure:  Azure pricing is easier to understand. It provides a dashboard with a billing section where it is quite clear to see how much you are spending and on what. Azure offers a pricing calculator that makes it easy to estimate the costs of services and a robust enterprise budgeting system that lets you allocate costs across departments or business units.

• Google Cloud:  In terms of simple pricing criteria, Google stands out for its attractive and customer-friendly pricing structure. It tries to beat prices offered by other cloud service providers to win over business.

All three vendors have a free tier where you can try their services before you buy—each provider even offers a “free forever” tier where they provide a limited set of services on an ongoing basis.

AWS Pricing Models

Let’s review AWS pricing models in more detail.

On-Demand Pricing

The default AWS pricing model is pay-as-you-go, and you are charged for services based on actual usage per hour or per second. It's flexible but is also the most expensive option. Many organizations start with on-demand pricing and then switch to other models as they better understand their cloud needs.

Reserved Instances

Amazon lets you pre-order instances for a term of 1 or 3 years and get up to 75% off the on-demand rate. In the Reserved Instance model, you cannot delete Reserved Instances when you need to scale them down (however, you can sell Reserved Instances on a dedicated Amazon Marketplace). To scale up, you will need to use more expensive on-demand resources. 

While this reduces the flexibility of Amazon services, you can still benefit from the advanced automation options and rich ecosystem of services that Amazon offers. Most organizations combine models, using Reserved Instances for long-running workloads and on-demand for workloads that experience fluctuations.

Spot Instances

Spot Instances are available on Amazon EC2, Amazon Fargate, and several other compute services. It offers the best discount with up to 90% off the on-demand instance price. Spot Instances allow you to bid on reserve computing power on Amazon's open market. The price changes every 5 minutes, and if your bid is above the current market price, you will receive a Spot Instance. 

The catch is that Spot Instances are terminated with notice of only 2 minutes when capacity is unavailable or the current Spot price exceeds the maximum price. Amazon has a new capability that lets you receive an advanced warning that a Spot Instance will be terminated, but this is not guaranteed. It also offers Spot Fleet, an advanced mechanism that manages scalability for groups of Spot Instances and regular on-demand instances.

Azure Pricing Models

Beyond its on-demand pricing model, Azure offers two primary cost optimization strategies: Azure VMs and Spot VMs.

Pay-As-You-Go

Azure services are billed per second based on actual usage, with no long-term commitments or upfront costs. This gives you the flexibility to increase or decrease resources as needed. Azure virtual machines (VMs) can be automatically resized using the Azure autoscale feature.

This pricing model is primarily intended for users who like flexibility and want to convert capital expenses into operating expenses, and applications with variable or short-term workloads.

Reserved Virtual Machine Instances (RVMIs)

Azure RVMIs are pre-ordered virtual machines which require a commitment of 1 or 3 years in select regions. RVMIs give you up to 72% off pay-as-you-go pricing.

Azure may choose to replace an RVMI with another instance during the term. Users can also cancel Reserved Instances before the end of the term, but this will incur an early termination fee.

This pricing model is suitable for applications with consistently stable load, organizations with a fixed budget, or large applications that constantly use a certain number of virtual machines (such as a central management component).

Spot VMs

Azure allows you to purchase unused compute power at up to a 90% discount compared to pay-as-you-go. However, Spot Instances can be interrupted by sudden notifications and are only considered suitable for workloads that can tolerate interruptions. The notice for spot VM interruption is only 30 seconds.

Azure provides Virtual Machine Scale Sets (VMSS). This is an autoscaling mechanism that allows you to manage groups of VMs and automatically add Spot Instances based on predefined policies. Unlike Amazon’s Spot Fleet, VMSS do not allow mixing Spot VMs and pay as you go VMs.

Spot Instances are primarily suitable for distributed fault-tolerant applications, stateless applications, and non-emergency or highly parallelized workloads.

Google Cloud Pricing Models

Google Cloud Platform offers the following pricing models:

Pay-As-You-Go

Google Cloud offers pay-as-you-go pricing. This is ideal for individuals looking to use the cloud intermittently as it allows the flexibility to add and remove services as needed. This level of flexibility comes at a price, so the pay-as-you-go model has the highest hourly cost on the platform.

Long-Term Commitment Plans

If you have a long-term cloud deployment plan and can make a long-term commitment, you can realize significant cost savings over a pay-as-you-go model. Google offers a long-term pricing model with a choice of 1 or 3 years in advance. Google named its plans Committed Use and offers discounts of up to 70% compared to the on-demand price.

Preemptible Instances

Preemptible VM instances (spot instances) are significantly cheaper (60-91% off) compared to standard VM prices. However, if the compute engine needs to reclaim compute capacity to allocate to other virtual machines, it can stop (preempt) those instances, with a notice of 30 seconds. Preemptible instances use excess Compute Engine capacity, so availability depends on usage.

Preemptible instances can significantly reduce Compute Engine costs if your application is fault-tolerant and can tolerate instance preemption. For example, batch jobs can run on preemptible instances. If some of these instances are stopped, the work will be slowed down, but will not complete. Preemptible instances complete batch jobs without placing additional workloads on existing instances or paying the full price for additional regular instances.

However, Google Preemptible Instances offer much less mature management capabilities compared to AWS and Azure, making it more difficult to scale and combine them with pay-as-you-go instances.

Conclusion

In this article, I explained the key differences between AWS, Azure, and Google Cloud in terms of pricing models. I focused on three primary payment tiers:

• On-demand/pay-as-you-go: Allows you to purchase cloud resources and pay per hour or minute of usage

• Reserved Instances/VMs: Allows you to commit to computing resources for a term of 1 or 3 years and receive discounts up to 72% (depending on the provider)

• Spot/Preemptible instances/VMs: Lets you buy spare capacity on the cloud provider’s spot market at discounts of up to 90%, but face the risk that instances will be terminated on short notice

I hope this will be useful as you evaluate your use of the public cloud in 2022 and beyond.