EC2 Spot Instances vs. AWS Savings Plans: What are the Potential Savings?

Some of our customers have been asking what the potential savings benefits are for using EC2 Spot Instances over AWS Savings Plans, so I wanted to take some time to address that here. 

EC2 Spot instances are discounted up to 90%, depending on which EC2 instance type, AWS region, and Availability Zone you select. On average, you’ll find savings between 50 and 70%. Spot instances are billed on a per-second basis, with a minimum of one minute. Keep in mind that per-second billing applies to EC2 instances running free editions of Linux, such as Amazon Linux 2, CentOS, or Ubuntu. If you’re running EC2 Spot Instances with a Microsoft Windows operating system, you are billed hourly, but still benefit from deep discounts on the hourly rate.

The caveat to EC2 Spot Instances is that AWS can reclaim the capacity with a two-minute warning. If you have workloads that aren’t critical or only need to scale up on a temporary basis, EC2 Spot Instances are an excellent option. 

In contrast, AWS Savings Plans share many similarities with Reserved Instances, or just “RIs.” Like RIs, Savings Plans require that you commit to a 1-year or 3-year period. AWS Savings Plans are an excellent choice if you plan to utilize a baseline amount of compute performance over a longer period of time. While AWS Savings Plans offer up to 72% savings on your AWS compute costs, the average range of savings is between 20 and 50%. 

There are two types of savings plans available: EC2 Instance and Compute. While the EC2 Instance Savings Plans provide slightly higher discounts, Compute Savings Plans give you the most flexibility with compute consumption, including EC2, Fargate, and Lambda (serverless).

Given these two options, which should you choose? I would recommend leaning towards EC2 Spot Instances whenever possible, as there is no time commitment, and the discounts are significant. If your application is not able to handle unexpected downtime, stores state locally, or is otherwise mission critical, then you should lean towards AWS Savings Plans.

FAQ

How do the management and operational complexities of using EC2 Spot Instances compare to AWS Savings Plans?
Managing EC2 Spot Instances can be more complex than utilizing AWS Savings Plans due to the nature of Spot Instances being subject to market availability and price volatility. Spot Instances can be interrupted with a two-minute notice, requiring a more robust architecture to handle such interruptions. This might involve implementing auto-scaling groups, spot fleet, or other AWS services to ensure high availability and fault tolerance. On the other hand, AWS Savings Plans offer a simpler management approach, where you commit to a consistent amount of usage (measured in $/hour) for a 1 or 3-year term in exchange for lower rates. There's no need to manage instance interruptions or availability, making Savings Plans easier to manage but less flexible than Spot Instances.

Can EC2 Spot Instances and AWS Savings Plans be used in conjunction to further optimize cost savings?
Combining EC2 Spot Instances and AWS Savings Plans can further optimize cost savings. While Savings Plans offer lower prices for committed use, leveraging Spot Instances for non-critical or flexible workloads can reduce costs even more. Organizations can balance cost and performance effectively by strategically using Spot Instances for suitable workloads and covering more predictable, steady-state workloads with Savings Plans. However, this approach requires a nuanced understanding of application workloads and AWS pricing models to ensure cost-effectiveness without compromising performance or availability.

Are there specific case studies or examples illustrating the savings achieved by companies using EC2 Spot Instances versus AWS Savings Plans?
While specific case studies were not mentioned in the blog post, many organizations have shared their experiences online, detailing significant cost savings by leveraging EC2 Spot Instances, AWS Savings Plans, or a combination of both. These case studies often highlight scenarios where companies have optimized their cloud expenditure by aligning their workload characteristics with the most suitable pricing model, showcasing strategies such as using Spot Instances for stateless, batch, or background tasks and Savings Plans for baseline, steady-state usage. These real-world examples can provide valuable insights into best practices and potential savings, although each company's results can vary based on their unique workload patterns and application architectures.