Streamlining bulky applications

Using machine learning to deliver next-generation web experiences

Peter Blum
Machine learning illustration via Shutterstock

The amount of JavaScript being used on websites has tripled in the past three years. Machine learning is what web developers need to speed up their increasingly bulky applications, says product management VP Peter Blum.

There’s been a massive increase in the use of HTML and JavaScript code as developers work to deliver immersive web experiences complete with responsive design, device optimization and native-like application experiences within standard browsers.

In fact, the amount of JavaScript used by the top 100 websites has almost tripled in the last three years, according to httparchive. All this extra code comes at a cost — it slows performance and compromises the user experience. To deter these negative effects, developers are looking to machine learning to streamline the delivery of bulky applications.

Where’s the bottleneck?

When a site loads and a browser request is made, traditional web delivery approaches need to send all of the JavaScript code to the end user’s browser. But typically more than half of this code is never even used. Yet it’s still all sent to the browser, choking the delivery process and slowing performance.

As browsers parse the extra code, deciding what is necessary and what is extraneous, users sit for seconds staring at blank screens as they wait for applications to load. In today’s zero-patience economy, each millisecond added to site load times is crucial. If web sites and mobile applications are too slow, even by 250 milliseconds, users will lose interest and move on, resulting in lost revenue for businesses.

To date, traditional web delivery approaches to improve JavaScript performance have been incremental. For example, “minification” removes line breaks, whitespace and comments from code, yet only provides baseline benefits.

Enter machine learning and the application architecture of the future

Imagine if browsers could intelligently decide what JavaScript code is actually used and only download that code on demand while ignoring dead, unused code. This is exactly what a new technology called SmartSequence from Instart Logic does. This technology applies machine learning to a wealth of JavaScript execution information sent by real users’ browsers to gain an understanding of how browsers use code. It then applies this information to optimize delivery and enhance performance for end users.

The system is able to detect what JavaScript code is used and only delivers the necessary portions up front and can demand-load other JavaScript as the application needs it. The system continually learns as the JavaScript changes over time and as consumer usage patterns evolve.

This novel approach to JavaScript delivery results in the reduction of download sizes for typical web applications by 30 to 40 percent. Higher performance rates mean increased conversions, reduced bandwidth costs and exceptional user experiences across devices. Armed with machine learning, developers can tackle the excess code problem, enhance application performance and push the boundaries on web development.

Peter Blum
Peter Blum is VP of Product Management at Instart Logic, helping revolutionize the way Internet gets delivered.

Inline Feedbacks
View all comments