Since 2019, AWS has introduced cost optimizations for both server-based computing with Amazon
Elastic Compute Cloud (Amazon EC2) and serverless technologies, including Lambda and Amazon
Elastic Container Service (Amazon ECS) with AWS Fargate, the serverless compute engine built on
Amazon ECS, and others.
New cost optimizations for Lambda
1. Lambda is now included in the Compute Savings Plans, a flexible pricing model that allows
customers to save up to 17 percent in exchange for making a commitment to a consistent
amount of compute usage ($10 hour, for example) for a one-year or three-year term.
2. Lambda duration billing changed granularity from 100 ms down to 1 ms, which lowers the cost
for most Lambda functions. This is especially pronounced for short-duration Lambda functions,
where customers can save up to 70 percent.
3. Lambda now supports function sizes up to 10 gigabytes of RAM and 6 vCPU, allowing customers
to reduce costs for resource-constrained, compute-intensive workloads.
4. Lambda AVX2 support allows customers to save up to 30 percent on compute-intensive,
vectorizable workloads.
5. Lambda is now included in the AWS Compute Optimizer, which allows customers to easily
identify and remediate inefficient configurations.
6. Lambda cost can be reduced with event filtering by configuring the event source with a filter
criterion that reduces the number of messages that are used to invoke the Lambda function.
7. Finally, with tiered pricing (based on compute duration measured in gigabytes-seconds),
customers who run large workloads on Lambda can automatically save on their monthly costs.
So customers with an x86 architecture for a 6–15 billion compute duration (gigabytes-seconds)
get a 10 percent discount, and for over 15 billion, a 20 percent discount. Similarly, for an arm64
architecture for compute duration (gigabytes-seconds) in the 7.5–18.75 billion range, customers
get a 10 percent discount, and for over 18.75 billion, a 20 percent discount.
3
