Serverless — Functions Without Servers

Run code in response to events — no server management, automatic scaling, pay per invocation.

When to use

• Event-driven workloads (webhooks, async processing, scheduled jobs)
• Spiky or infrequent traffic (scales to zero, no idle cost)
• Background processing that does not need low latency

When NOT to use

• Long-running processes (>15 min timeout on most platforms)
• Latency-sensitive hot paths (cold starts can be 100ms–3s)
• Stateful workloads

Tradeoffs

• Cold starts: significant for JVM, moderate for Node/Python, minimal for Go
• Vendor lock-in on trigger types, runtime config, and resource limits

Go

package main

import (
    "context"
    "fmt"
    "github.com/aws/aws-lambda-go/events"
    "github.com/aws/aws-lambda-go/lambda"
)

func handler(ctx context.Context, req events.APIGatewayProxyRequest) (events.APIGatewayProxyResponse, error) {
    userID := req.QueryStringParameters["user_id"]
    if userID == "" {
        return events.APIGatewayProxyResponse{StatusCode: 400, Body: "missing user_id"}, nil
    }
    result := fmt.Sprintf(`{"user_id": "%s", "status": "ok"}`, userID)
    return events.APIGatewayProxyResponse{StatusCode: 200, Body: result}, nil
}

func main() { lambda.Start(handler) }

Python

import json

def handler(event: dict, context) -> dict:
    user_id = event.get("queryStringParameters", {}).get("user_id")
    if not user_id:
        return {"statusCode": 400, "body": "missing user_id"}

    # perform work — call DB, publish event, etc.
    result = {"user_id": user_id, "status": "ok"}
    return {
        "statusCode": 200,
        "body": json.dumps(result),
    }

Gotcha: Cold starts are worst with large runtimes and large deployment packages. Go has the fastest cold starts (~50ms). JVM is worst (~3s). Use provisioned concurrency for latency-critical serverless endpoints.