Fat Lambda — Everything You Need to Know

Slow Lambda Cold Start / Boot

It takes longer than usual for a Lambda function to process the initial request. Why?

Lambda Throttling

Single Point of Failure

Having multiple Lambda functions to handle individual requests reduces single point of failure, because all your code is not in the same function.

You make one mistake in setting up your Lambda function; all your APIs go down, not just one. Examples:

• Forgetting to add environment variables
• Top level await which throws an error
• Insufficient IAM permission provided to Lambda role

This way, if one Lambda function goes down → None of your APIs will work, resulting in status 5xx server errors.

Violates Least Privilege Policy

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "BAD CODE - Do not copy!",
      "Action": "s3:*",
      "Effect": "Allow",
      "Resource": "*"
    }
  ]
}

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "Stmt1706106939799",
      "Action": ["s3:GetObject"],
      "Effect": "Allow",
      "Resource": "arn:aws:s3:::BUCKET_NAME/*.csv"
    }
  ]
}

Poor CloudWatch Logging / Debugging

In this case, the log of a particular route won’t be easy to find in CloudWatch logs, because everything is being logged to a single CloudWatch log group.

It becomes tedious to filter the logs of API Gateway routes as well, because here the sole purpose of API Gateway becomes to just invoke the Lambda function.

Poor Lambda CloudWatch Metrics

Since all APIs exist in a single Lambda, you can’t set CloudWatch alarms on particular routes, or monitor the performance of the AWS Lambda Function.

Using API Gateway just to invoke the Lambda function

AWS API Gateway doesn’t get used to its full potential, because the router is built into the fat Lambda function.

We should be using API Gateway routes to map our routes with different Lambda function which can be seen below.

Fat Lambda optimization workout plan:

• Start by ditching the request router from your Lambda e.g. Express, Django. API Gateway exist.
• Create multiple Lambdas with IaC like CDK, Serverless framework or SST
• Use Amazon API Gateway to route your requests
• Map requests in API Gateway based on the method and route parameters

Middleware - The traffic controller

Every framework like ExpressJS gives you a handy-dandy middleware but how do we get them on Lambda?

I hate manually writing code to validate every request!

I bring to you MiddyJS — the stylish NodeJS middleware engine for AWS Lambda.

Here is a simple example from their website:

import middy from "@middy/core";

const lambdaHandler = (event, context) => {
  // REAL BUSINESS LOGIC
  return doSomethingUsefulWith(event);
};

export const handler = middy(lambdaHandler)
  .use(/* Your own behaviour in a reusable fashion */)
  .use(/* logging and tracing */)
  .use(/* input validation */)
  .use(/* authentication */)
  .use(/* error handling */)
  .use(/* other behaviour */);