AWS Lambda performance tuning

Introduction

The genesis for this solution was based on a Chalk Talk for re:invent 2023 - SVS319 AWS Lambda performance tuning: Best practices and guidance. Our goal is to take an already good API and make it great.

Our focus is to create an optimized solution for a maximum number of transactions per second (TPS) using Lambda.

The solution idea/starting point was from Serverless Land which was a very basic solution for reading/writing records from an Aurora RDS. Using the very popular ORM (Object Relational Mapping) sequelize library, allowed Lambda to connect to Aurora RDS via RDS Proxy, write a list of NFL stadiums from a JSON file or read from the datbase. This was all invoked by an Amazon API Gatweay (HTTP), and deployed via CDK as Infrastrucure as Code (IaaC). We refer to this as the unoptimized soltuion. This solution on average took ~250ms to execute. After making a few key changes to update the solution, optmized solution, where the execution time is now ~9ms.

More Details

if you are interested to learn more about the thought process behind the decisions listed below, we highly encourage to read the related blog post to this example

Pre-requisites

• Node.js
• AWS CLI
• AWS CDK

Deploying the solutions

Wnat to deploy this for yourself? With CDK, this is very easy. You will find two diretories, optimized and unoptimized. You'll need to deploy each solution seprately.

Deployment commands

• cdk bootstrap bootstrap your environment with CDK
• cdk deploy --all deploy this stack to your default AWS account/region

Unoptimized Solution

The unoptimized solution, contains the following:

• sequlize ORM library for conecting to PostgresSQL (Amazon Aurora Postgres)
• AWS SDK
• RDS Proxy
• Runtime: NodeJS 18 with TypeScript
• Parameter Store for storing parameters
• Calling AWS SDK during runtime
• x86 architecture
• 256MB memory size

Optimized Solution

The optimized solution, contains the following:

• Postgres.js library for conecting to PostgresSQL (Amazon Aurora Postgres)
• RDS Proxy
• Runtime: NodeJS 18 with TypeScript
• Parameter Store for storing parameters
• Retrieving parameters through a Lambda extension
• Amazon ElastiCache
• Graviton 2 architecture
• 1024MB memory size

Deleting the Solution

To delete the stacks via AWS Console follow these instructions:

1. Open the CloudFormation Console page and choose the relevant stack, then choose "Delete"
2. Once the confirmation modal appears, choose "Delete stack".
3. Wait for the CloudFormation stack to finish updating. Completion is indicated when the "Stack status" is "DELETE_COMPLETE"

License

This library is licensed under the MIT-0 License. See the LICENSE file.