Serverless Functions for Website (AWS Lambda)

Developing Serverless Functions for Website (AWS Lambda)

AWS Lambda — functions that run on event and don't need constantly running server. Pay only for actual execution time: first 1M requests monthly free, then $0.20 per 1M requests. Cold start is the only serious drawback, solved several ways.

When Lambda Fits for Website

Fits: contact form processing, PDF/image generation on request, webhooks from payment systems and CRM, image resizing on upload, task schedulers (cron via EventBridge), API proxy for third-party services.

Doesn't fit: long-lived connections (WebSocket needs separate service), tasks > 15 minutes, high-frequency DB operations without connection pooling.

Project Structure

my-lambda/
├── src/
│   ├── handlers/
│   │   ├── contact-form.ts
│   │   ├── image-resize.ts
│   │   └── webhook.ts
│   ├── services/
│   │   ├── email.ts
│   │   └── storage.ts
│   └── utils/
├── template.yaml          # SAM or serverless.yml
├── package.json
└── tsconfig.json

Example Function: Contact Form Processing

import { APIGatewayProxyHandler } from "aws-lambda";
import { SESClient, SendEmailCommand } from "@aws-sdk/client-ses";
import { z } from "zod";

const ses = new SESClient({ region: "eu-west-1" });

const ContactSchema = z.object({
  name: z.string().min(2).max(100),
  email: z.string().email(),
  message: z.string().min(10).max(2000),
});

export const handler: APIGatewayProxyHandler = async (event) => {
  const headers = {
    "Access-Control-Allow-Origin": "https://your-site.com",
    "Content-Type": "application/json",
  };

  try {
    const body = JSON.parse(event.body || "{}");
    const data = ContactSchema.parse(body);

    await ses.send(new SendEmailCommand({
      Source: "[email protected]",
      Destination: { ToAddresses: ["[email protected]"] },
      Message: {
        Subject: { Data: `New message from ${data.name}` },
        Body: {
          Text: { Data: `From: ${data.name} <${data.email}>\n\n${data.message}` }
        }
      }
    }));

    return { statusCode: 200, headers, body: JSON.stringify({ ok: true }) };

  } catch (error) {
    if (error instanceof z.ZodError) {
      return { statusCode: 400, headers, body: JSON.stringify({ errors: error.errors }) };
    }
    console.error(error);
    return { statusCode: 500, headers, body: JSON.stringify({ error: "Internal error" }) };
  }
};

Deploy via AWS SAM

template.yaml:

AWSTemplateFormatVersion: "2010-09-09"
Transform: AWS::Serverless-2016-10-31

Globals:
  Function:
    Runtime: nodejs20.x
    Timeout: 10
    MemorySize: 256
    Environment:
      Variables:
        NODE_ENV: production

Resources:
  ContactFormFunction:
    Type: AWS::Serverless::Function
    Properties:
      Handler: dist/handlers/contact-form.handler
      Events:
        Api:
          Type: HttpApi
          Properties:
            Path: /contact
            Method: POST
      Policies:
        - SESCrudPolicy:
            IdentityName: your-site.com

npm run build
sam build
sam deploy --guided  # first deploy
sam deploy           # subsequent

Cold Start and Mitigation

Cold start on Node.js 20 — 200–500 ms. Python — 100–300 ms. After first call function stays "warm" minutes.

Mitigation ways:

• Provisioned Concurrency — Lambda keeps N instances always warm. Costs ~$0.015 per hour per instance. For critical endpoints with SLA.
• Reduce bundle size — use esbuild instead webpack, don't include aws-sdk in bundle (already in environment), tree-shaking.
• SnapStart (Java) — snapshot state after initialization.
• Warmer — EventBridge calls function every 5 minutes with warmup-payload.

// Initialization outside handler — executes once on cold start
const dbClient = new DynamoDBClient({ region: "eu-west-1" });
const sesClient = new SESClient({ region: "eu-west-1" });

export const handler = async (event) => {
  // handler uses already initialized clients
};

Working with Database

Standard TCP connection to PostgreSQL/MySQL doesn't fit Lambda — each call creates new connection, at 1000 RPS database chokes.

Solutions:

• RDS Proxy — connection pool before RDS, Lambda connects to proxy. +$0.015 per vCPU/hour RDS.
• DynamoDB — no connections, natively serverless. For simple data structures.
• PlanetScale / Neon — serverless-compatible databases with HTTP API.

Logging and Tracing

import { Logger } from "@aws-lambda-powertools/logger";
import { Tracer } from "@aws-lambda-powertools/tracer";

const logger = new Logger({ serviceName: "contact-form" });
const tracer = new Tracer({ serviceName: "contact-form" });

export const handler = tracer.captureLambdaHandler(async (event) => {
  logger.addContext(context);
  logger.info("Processing contact form", { email: event.body?.email });
  // ...
});

AWS Lambda Powertools — official library for structured logging, tracing via X-Ray, metrics via CloudWatch EMF.

Timeline

One Lambda function with SAM deploy — 1–2 days. Set of 5–7 functions with CI/CD via GitHub Actions and monitoring — 5–7 days.