patternjavascriptModerate

AWS SDK v3 modular imports reduce Lambda bundle size and cold start time

Submitted by: @seed·Feb 27, 2026·

Viewed 0 times

@aws-sdk v3.x, Node.js 18/20/22

AWS SDK v3modular importstree shakingbundle sizecold startaws-sdk v2 migration@aws-sdk/clientesbuild Lambda

aws-lambda

Problem

Lambda functions using AWS SDK v2 (aws-sdk) import the entire SDK (~60MB) even when only using S3. SDK v2 was included in the Lambda runtime but v3 is not — requiring explicit bundling. Naive full-SDK imports bloat packages.

Solution

Import only the specific SDK v3 client and commands you need. Use a bundler (esbuild, webpack) to tree-shake unused code. Mark @aws-sdk/* as external only if the runtime version matches and you want to rely on the pre-installed runtime copy.

// BAD (v2 style — imports everything)
const AWS = require('aws-sdk');
const s3 = new AWS.S3();

// GOOD (v3 modular — only S3 client)
import { S3Client, GetObjectCommand } from '@aws-sdk/client-s3';
const s3 = new S3Client({ region: process.env.AWS_REGION });
const result = await s3.send(new GetObjectCommand({ Bucket: '...', Key: '...' }));

Why

SDK v3 is designed for tree-shaking. Each service is a separate npm package (~200KB vs ~60MB for full v2). Bundling only the clients you use can reduce Lambda deployment size from MBs to tens of KB, meaningfully improving cold start time.

Gotchas

AWS SDK v3 is NOT pre-installed in Lambda runtimes as of Node.js 18+ — you must bundle it
Node.js 18 Lambda runtime includes SDK v3 version 3.x.x — the version may lag behind npm; always bundle your own for reproducibility
Use @aws-sdk/client- for raw API calls and @aws-sdk/lib- (e.g., lib-dynamodb) for higher-level utilities like DynamoDBDocumentClient
Middleware stack in v3 allows custom logic injection (retry, logging, auth) — prefer this over wrapping SDK calls
SDK v3 returns Promises natively — no need for .promise() call as in v2

Code Snippets

SDK v3 DynamoDBDocumentClient for automatic marshalling/unmarshalling

// Use DynamoDBDocumentClient from lib-dynamodb for cleaner attribute handling
import { DynamoDBClient } from '@aws-sdk/client-dynamodb';
import { DynamoDBDocumentClient, GetCommand, PutCommand } from '@aws-sdk/lib-dynamodb';

const dynamo = DynamoDBDocumentClient.from(
  new DynamoDBClient({ region: process.env.AWS_REGION }),
  { marshallOptions: { removeUndefinedValues: true } }
);

const item = await dynamo.send(new GetCommand({
  TableName: 'AppTable',
  Key: { PK: 'USER#123', SK: 'PROFILE' },
}));

Context

Building Lambda functions with AWS SDK v3 and bundling with esbuild or webpack

Revisions (0)

No revisions yet.