AWS Lambda#

So you’ve got a simple application you want to run in the cloud. Maybe it’s an API endpoint, a scheduled task, or a function that processes files uploaded to Amazon S3. You could spin up an EC2 instance, set up an ECS cluster, or run code directly on a server.

AWS Lambda is a serverless compute service. You write your code and upload it, and AWS handles the rest: provisioning servers, scaling, patching, and monitoring. You only pay for the compute time you actually use.

The Serverless Pitch#

Here’s the thing: with EC2 or ECS, you’re paying for servers whether they’re busy or idle. If your application only handles 100 requests a day, you’re still paying for a server to sit there 24/7 waiting for those requests.

Lambda flips this model. Your code only runs when it’s triggered, and you only pay for those milliseconds of execution time. No requests? No charges (beyond a tiny storage fee for the code itself).

When Lambda Makes Sense#

  • Event-driven workloads: Process S3 uploads, respond to SNS messages, react to DynamoDB changes
  • APIs with variable traffic: Handle thousands of requests per second or just a handful per day
  • Scheduled tasks: Run a job every hour without maintaining a server
  • Background processing: Resize images, generate PDFs, send emails
  • Prototypes and MVPs: Get something running fast without infrastructure setup

When Lambda Does Not Make Sense#

  • Long-running processes: Lambda functions time out after 15 minutes max
  • Stateful applications: Each invocation is isolated; you can’t keep state in memory between requests
  • Applications needing large dependencies: 250 MB unzipped deployment package limit
  • Consistent high traffic: If your function runs constantly, EC2 or ECS might be cheaper
  • Applications requiring specific OS customization: You get a standard runtime environment

How Lambda Works#

When a Lambda function is invoked:

  1. Cold start (first time or after idle): AWS provisions a container, loads your code, and starts the runtime. This adds latency (often a few hundred milliseconds for Python, longer for Java or .NET).
  2. Warm execution (subsequent requests): If another request comes in quickly, AWS reuses the existing container. This is fast (single-digit milliseconds).
  3. Scaling: If requests come in faster than the available containers can handle, AWS automatically creates more containers. Lambda scales up to thousands of concurrent executions.
Backward lambda deploy-lambda Forward