Serverless computing

1. What is serverless? #

The ultimate level of abstraction, where the server becomes invisible.

No-Ops? #

Serverless does not mean there are no more servers, but that you no longer have to manage them. You only provide a function (FaaS) that runs in response to an event (click, file upload, message in a queue).

2. The advantages: cost and agility #

Cost is calculated by the number of executions and their duration (often to the millisecond). If your code does not run, you pay €0. This enables instant scaling from 1 to 10,000 requests per second.

3. Use cases and limitations #

Ideal for APIs, image processing or scheduled tasks. Watch out, however, for 'cold starts' (latency on first start) and execution time limits (often 15 minutes max per function).

Hands-on: a serverless function #

Deploying and calling a function.

Writing and deploying #

# handler.py
def handler(event, context):
    return {"statusCode": 200, "body": "Bonjour depuis le serverless"}

# AWS Lambda
zip function.zip handler.py
aws lambda create-function --function-name bonjour --runtime python3.12 --handler handler.handler \
  --zip-file fileb://function.zip --role arn:aws:iam::123456789012:role/lambda-role
aws lambda invoke --function-name bonjour out.json && cat out.json
aws logs tail /aws/lambda/bonjour --follow

# Google Cloud Functions / Azure Functions
gcloud functions deploy bonjour --gen2 --runtime=python312 --trigger-http --allow-unauthenticated --region=europe-west1
func init MonProjet --python && func new --name bonjour --template "HTTP trigger" && func start

Common mistake: Cold starts, duration limits and vendor lock-in are the three pitfalls of serverless: measure the latency of the first call.