This repository serves as a companion to my talk at AWS ReInvent. The talk, titled "“Rustifying” serverless: Boost AWS Lambda performance with Rust", explores how to leverage the performance advantages of Rust in serverless AWS Lambda functions. For more details and to watch the talk, visit ReInvent Talk.

• As a company focused on developing AWS management tools, we started with a Minimum Viable Product (MVP) for S3 bucket management.
• Initially, the product was open only to our most loyal customers, and we received overwhelmingly positive feedback.
• Eventually, we opened the product to all of our users.
• At its peak, the service catered to thousands of clients and tens of thousands of users.

• We started experiencing performance bottlenecks and increased IT spending.
• Major architectural changes like adding cache and improving algorithms were implemented.

• We reached a point where our existing runtime, Python, was not sufficient for our performance needs.
• Thus, we decided to leverage Rust for its performance advantages.

This repository illustrates the journey that led us to this point.

• The architecture consists of an API Gateway with a Lambda authorizer.
• Multiple Lambda functions are connected to the API Gateway. These functions interact with services like S3 and DynamoDB.
• We have multiple runtimes, mostly Python, but also NodeJS.

There are primarily three ways to integrate Rust:

• Use Pyo3 with maturin to create a .whl package for use in your Python Lambda.
• Example code can be found under s3-ops-rust-pyo3-lib.

• Use NAPI-RS to package and build your code. A one stop CLI, unlike Python where you need to use teo tools.
• Examples can be found under s3-ops-rust-napi-lib.

Benefits: Speed of development and no need to rewrite the Lambda function.

• Use cargo-lambda and AWS SAM to deploy a full-fledged Rust Lambda.
• Example code can be found under s3-admin-app/authorizer_rust.
• Benefits: Reduced cold starts, thanks to the Rust runtime.

• Use extensions to address cross-cutting concerns, such as analytics reporting, using a fast language like Rust.
• Example code can be found under analytics-extension.
• Benefits: Reduce Lambda latency by offloading tasks to an external process.

• This example comes with a devcontainer for both Rust and Python development.
• In case you are not using DevContainers, make sure you have the following prerequisites:
  • Python 3.11
  • Rust - latest version
  • Poetry
  • Poe
  • AWS SAM
  • NodeJS 20
  • AWS CLI
  • NAPI-RS
• Prepare the NodeJS Rust Library by running:

cd s3-ops-rust-napi-lib
npm i

• We use poetry for build management. To build and deploy, run:

  # Install dev tools used for rust compilation.
  poetry install --only=rust-dev-tools
  # Build the rust package
  poe build-python-lib
  poe build-node-lib
  # Build and deploy the extension
  poe build-and-deploy-extension

• Update s3-asmin-app/template.yaml with the ARN of the extension under Globals/Layers.
• Build the application and deploy it - poe build-and-deploy-app
• To create a new user in the 'users' table, follow the AWS SAM output. For example aws dynamodb put-item --table-name ...
• Make API calls to the various APIs as defined in the AWS SAM output.

The main application resides in s3-admin-app, Rust bindings are in s3-ops-rust-lib and the extension in analytics-extension.

 # Install dev tools used for rust compilation.
 poetry install --only=rust-dev-tools
 # Build the rust package
 poe build-python-lib
 poe build-node-lib
 # Add the local rust package
 poetry add .rust-lib/s3_ops_rust-0.1.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl --group dev 
 poetry install