Serverless is good, serverless is cool, serverless is innovative. But how far can we push with its uses? Let’s navigate some ups and downs around serverless technology.

When someone hears and "understands" the serverless concept for their fist time. what come to their mind is, So: I can move all these my applications to serverless and save cost?

• I can have managed services but yet very cheap?
• This mean no more paying for idle servers?
• This this remove the need for DevOps?
• I can scale automatically without any complex setup?

Stop being happy (^_-), you are assuming without knowing what is going on. The right questions comes when understanding: What, Why, When, How, Who regarding serverless.

What is Serverless ?

1. Execution Model

Your code runs:

• On demand
• In response to events (HTTP requests, messages, schedules, file uploads, etc.)
• For short-lived executions (typically seconds to minutes) You don’t keep servers running, the platform spins up execution environments when needed.

2. Cost Model

You pay for	Great for	Not always cheaper for
Execution time	Spiky traffic	Constant high-throughput workloads
Number of requests/events	Infrequent workloads	Long-running processes
Resources consumed (memory, duration)	Event-driven systems	Predictable, steady traffic

3. Scaling Model

What scaling is	What scaling is not
Automatic	You still need to understand concurrency limits
Horizontal	You still design for backpressure and failures
Event-driven	Bad architecture can still scale expensive problems

4. Operations Reality

Serverless does not remove DevOps, it changes it.

What you still need	Where the difference lays
Monitoring and logging	Less infrastructure ops
CI/CD pipelines	More platform
Security and IAM	More observability
Cost controls	More architecture work
Incident response	More studies of compatibilities

The Right Questions Come After Understanding

Once the excitement fades, the real questions begin: What workloads fit serverless?

• Why should I use it instead of containers or VMs?
• When does it save cost and when does it not?
• How do I design, test, monitor, and debug at scale?
• Who owns reliability, security, and cost control?

Serverless is a powerful tool, not a universal solution.

Do I have your attention now? Fine, let’s code!

Steps to Migrate

Step 0: The Starting Point (FastAPI App)

# apps.py

from fastapi import FastAPI, status

app = FastAPI()

@app.get("/health", status_code=status.HTTP_200_OK)
def health():
return {"status": "ok"}

@app.post("/items", status_code=status.HTTP_201_CREATED)
def create_item():
return {"message": "item created"}

@app.put("/items/{item_id}", status_code=status.HTTP_200_OK)
def update_item(item_id: int):
return {"message": f"item {item_id} updated"}

Running the code:

uvicorn app:app --reload

Step 1: Understand the Serverless Mapping (Mental Model)

Step 2: Add Mangum (ASGI → Lambda Adapter)

Install dependencies

pip install fastapi mangum

Update to apps.py file.

from fastapi import FastAPI, status
from mangum import Mangum

app = FastAPI()

@app.get("/health", status_code=status.HTTP_200_OK)
def health():
return {"status": "ok"}

@app.post("/items", status_code=status.HTTP_201_CREATED)
def create_item():
return {"message": "item created"}

@app.put("/items/{item_id}", status_code=status.HTTP_200_OK)
def update_item(item_id: int):
return {"message": f"item {item_id} updated"}

# Lambda entry point
handler = Mangum(app)

The handleris what AWS Lambda will invoke.

Step 3: Create requirements.txt

fastapi
mangum

Step 4: Package the Application for Lambda

Create a build directory

mkdir package
pip install -r requirements.txt -t package
cp app.py package/

Zip it

cd package
zip -r app.zip .

Step 5: Create the Lambda Function

Go to Go to

1. AWS Lambda
2. Create functionAWS Lambda

• Runtime: Python 3.10+
• Architecture: x86_64

1. Upload app.zip
2. Set Handler: app.handler
3. Memory: 512 MB (good default)
4. Timeout: 10–15 seconds

Step 6: Create API Gateway (HTTP API)

1. Go to API Gateway
2. Create HTTP API
3. Integration:

• Type: Lambda
• Choose your Lambda function

1. Routes:

• ANY /{proxy+}

1. Deploy

This single route lets FastAPI handle all routing internally.

Step 7: Test your endpoint

Its a wrap!