Front End Optimization
for SPA

Trần Trọng Thanh

About me:

Trần Trọng Thanh

Front End Specialist

CTO & Co-founder
Front End Architect

Contact me:
Github, Twitter, Facebook: @trongthanh

Today's discussion:

How to make your web apps load fast

Mobile site users abandon after three seconds to load
“Amazon's calculated page load slowdown of just 01 second could cost it $1.6 billion in sales each year”

Common Advices

  • Concatenate JS & CSS
  • Minify static files
  • Optimize image sizes
  • Move <script> tag to bottom
  • Turn on gzip
  • ...

Let's try to answer:


What happen

from entering URL on address bar to full rendered page

HTTP Request


devtool network timing

DNS Lookup

Initial Connection

Time to First Byte (TTFB)

Advices for Initial Connection & TTFB

  • Reduce geographical latency with server location, CDN (devops)
  • Cache server requests (nginx, REDIS...) (backend + devops)
  • Reduce number of requests: concatenate, sprite sheet...
  • OR use HTTP2 (multiplexing) (devops)

Content Download

Advices for Content Download

  • Reduce static files size: minification
  • Optimize images, even better, resize images for difference screen size
  • Use gzip (backend + devops)
  • Use browser cache (backend + devops)
  • (2018) Use service-worker with cache API

Critical Rendering Path

<!doctype html>
		<meta charset="utf-8">
		<title>Page title</title>
		<link rel="stylesheet" href="/css/global.min.css">
		<link rel="stylesheet" href="/css/page.min.css">
		<script src="/js/vendor.min.js"></script>
		<script src="/js/main.min.js"></script>

		<!-- HTML markups here -->

critical rendering path

Advices for Critical Rendering Path

  • Inline critical (above the fold) CSS
  • Simplify HTML tags, avoid unnecessary nesting
  • Minify HTML to under 14kb (TCP packet limit)
  • Delay loading of non-critical assets: JS, below the fold CSS, web fonts...

Having done all of above will guarantee fast page render under 2 seconds.

Is it?

How about SPA?

SPA: Single Page Application

Typical SPA's downloaded HTML

<!doctype html>
<html lang="en">
		<meta charset="utf-8">
		<title>Single Page Application</title>
		<link rel="stylesheet" href="/app.min.css">
		<link rel="stylesheet" href="">

		<div id="app"></div>

		<script src="/bundle.min.js"></script>

Common performance problems with SPA:

  • Empty initial HTML
  • Large JS bundle size
  • Long JS parsing time
  • Critical rendering path doesn't apply for SPA

User-centric performance metrics

First Paint -> First Contentful Paint -> First Meaningful Paint -> Time to Interactive
FP -> FCP -> FMP -> TTI

SPA page speed optimization target:

Reduce time to FMP

Reduce JS bundle size

SPA Optimization Rules of Thumb

  • Minify and bundle JS on production
  • Shared bundles (leverage browser cache)
  • Code splitting (reduce load size for a "logical page")
  • Tree shaking
  • preload and prefetch

What is common among Angular, React, Vue?

Angular logo React logo Vue logo

Webpack flow cover


Src: React Vietnam Facebook Group

Minify JS on production

Use Webpack's UglifyJSPlugin

	config.plugins = config.plugins.concat([
		new webpack.optimize.UglifyJsPlugin({
			beautify: false,
			mangle: {
				screw_ie8: true,
				keep_fnames: true,
			compress: {
				screw_ie8: true,
				drop_debugger: true,
				drop_console: true,
			output: {
				comments: false,
			sourceMap: !PRODUCTION,

Shared Bundle:

Extract Style rules to CSS file

rules: {
	test: /\.s?css$/,
	use: ExtractTextPlugin.extract({
		fallback: 'style-loader',
		use: [{
				loader: 'css-loader',
				options: {/*...*/},
			}, {
				loader: 'postcss-loader',
				options: {/*...*/},
			}, {
				loader: 'sass-loader',
				options: {/*...*/},

plugins: [
	new ExtractTextPlugin('styles/styles-[chunkhash].css'),

Shared Bundle:

Extract common modules with DLLPlugin

const config = {
	entry: {
		vendor: [
	output: {
		filename: 'vendor.js',
		path: path.join(__dirname, buildPath),
		library: '[name]',
	plugins: [
		new webpack.DllPlugin({
			name: '[name]',
			path: `${buildPath}vendor-manifest.json`,

Code Splitting:

Split chunks based on route
Lazy load module with dynamic imports

<Route path="achievement" getComponent={(nextState, cb) => {

	import(/* webpackChunkName: "views/admin/achievement" */ './views/admin/AdminCoreValuePage' )
		.then(PageComp => {
			return cb(null, PageComp.default);

Lazy loaded route component with dynamic import

Tree Shaking

Webpack Bundle Analyzer

$ npm install --global webpack-bundle-analyzer

$ webpack --profile --json > stats.json

$ webpack-bundle-analyzer stats.json

Common tree shaking-ready techniques

  • Prefer multiple exports than single object export

// utils.js
// instead of this:
export default const utils = {
	trim() {/*...*/},
	format() {/*...*/},

// do this:
export function trim() {/*...*/};
export function format() {/*...*/};

Common tree shaking-ready techniques

  • Import only what needed, especially from multi-part libraries

// Ends up pulling in everything.
import { truncate } from 'lodash';

// Just pull in the module we're interested in.
import truncate from 'lodash-es/truncate';

Common tree shaking-ready techniques

  • Use Webpack IgnorePlugin (for example Moment)
    plugins: [
    	new webpack.IgnorePlugin(/^\.\/locale$/, /moment$/),
  • Or use tree-shakeable alternative: date-fns
    import {format, parse} from 'date-fns/esm';

preload and prefetch

<link rel="preload" href="/page/index.js" as="script"/>
<link rel="preload" href="/page/_error.js" as="script"/>
<link rel="preload" href="/app.js" as="script"/>

Preload common chunks and next page JS bundle for faster transition to next page

Wow, that's a lot!

But, there's more!

Can we make critical rendering path works for SPA?


With Universal JS Applications

What is Universal JS Applications

  • Used to be called Isomorphic JS Applications
  • There are portions of JS run on both client and server (hence universal / isomorphic)
  • Server-side rendered HTML
  • Server-enabled History Push State

Client only Redux architecture

Universal Redux architecture

That's all folks!
How do we audit our optimization effort?

Audit with LightHouse

Benchmark on production with

What we discussed was only part of
Front End Optimization

The whole picture: The RAIL model

To be continued...

Thank you!