aws-lambda-power-tuning VS failure-lambda

Sometimes, changing services, like opting for HTTP over REST API Gateway, leveraging tools like Lambda Powertuning to optimize functions, or reducing a CloudWatch log retention and changing log level, can lead to significant savings.

In case of not enabling SnapStart for the Lambda function we observed that increasing memory reduces the warm execution time for our use case especially for p>90. As adding more memory to the Lambda function is also a cost factor, the sweet spot between cold and warm start time and cost is somewhere between 768 and 1204 MB memory setting for the Lambda function for our use case. You can use AWS Lambda Power Tuning for very nice visualisations.

The aws lambda power tuning tool helps optimise the Lambda performance and cost in a data-driven manner. Let's try it out:

For Lambda, a very useful tool to help optimise is the AWS Lambda Power Tuning tool, released by Alex Casalboni, Developer Advocate at AWS: https://github.com/alexcasalboni/aws-lambda-power-tuning

Lambda memory affects not only the CPU performance and and host execution priority, but also network performance. Be wary though as the price scales linearly. You can use a tool like Lambda Power Tuning to find the sweet spot for your application. https://github.com/alexcasalboni/aws-lambda-power-tuning

This tool, which is open source and available here, takes the form of a Step Function that is deployed on your AWS account. The purpose of this Step Function is to run your lambda with different memory configurations several times and output a comparison in the form of a graph (or JSON) to try to find the optimal balance between cost and execution time. There are three possible optimization modes: cost, execution time, or a "balanced" mode where it tries to find a balance between the two.

The AWS Documentation's Memory and Computing Power page is a good starting point. To avoid configuring it manually, it's worth checking out AWS Lambda Power Tuning, which will help you find the sweet spot.

Utilizing tools such as AWS X-Ray, AWS Lambda Power Tuning, and AWS Lambda Powertools tracer utility is recommended. Read more about it here.

Install the AWS SAM CLI in your local environment. Configure your AWS credentials (requires AWS CLI installed): $ aws configure Clone this git repository: $ git clone https://github.com/alexcasalboni/aws-lambda-power-tuning.git Build the Lambda layer and any other dependencies (Docker is required): $ cd ./aws-lambda-power-tuning $ sam build -u sam build -u will run SAM build using a Docker container image that provides an environment similar to that which your function would run in. SAM build in-turn looks at your AWS SAM template file for information about Lambda functions and layers in this project. Once the build has completed you should see output that states Build Succeeded. If not there will be error messages providing guidance on what went wrong. Deploy the application using the SAM deploy "guided" mode: $ sam deploy -g

Use AWS Lambda Power Tuning to balance cost and performance.

failure-lambda is a small Node module for injecting failure into AWS Lambda. It offers a simple failure injection wrapper for your Lambda handler where you then can choose to inject failure by setting the failureMode to latency, exception, denylist, diskspace or statuscode. https://github.com/gunnargrosch/failure-lambda

In one of our last projects, I started using Failure-Lambda module from Gunnar Grosch to be able to inject failure in our system. Not yet in production, but it helped a lot during the test and staging phase to detect and handle possible errors that we overlooked. (If you want to know more about Chaos Engineering and Failure Lambda check out this post) So I was very eager to see his speech about Chaos Engineering at the Hamburg Serverless Days.

A common approach is to implement a circuit breaker inside your function and use DynamoDB to store the circuit breaker state (like Gunnar Grosch’s failure lambda node.js implementation and Jeremy Daly outlines in his AWS Reference Architecture Pattern). The Lambda function will fail before calling the Third-Party API when a failure threshold has been exceeded. This protects the downstream service, but it will not stop AWS Lambda polling the upstream queue and invoking your function. You also have to make changes to lambda function code, specific to the particular Lambda runtime and programming language. The approach introduces a number of DynamoDB requests which could significantly increase costs.