The Code Problem for Web Applications & How Instart Logic Is Using Machine Learning To Fix It

The adoption of mobile devices with powerful CPUs and full HTML5 browsers is enabling a new wave of increasingly sophisticated applications. Responsive/adaptive sites and single page applications are two obvious examples. But the increased sophistication is also creating new performance bottlenecks along the application delivery path. While the industry can continue to eek out incremental performance gains from network-level optimizations, the innovation focus has shifted to systems that are application-aware (like FEO) and now execution-aware. It’s the new frontier for accelerating application performance.

To deliver web experiences that meet these new world demands, developers are increasingly using JavaScript in web design. In fact, according to httparchive, the amount of JavaScript used by the top 100 websites has almost tripled in the last three years and the size of web pages has grown 15 percent since 2014. The popularity of easily available frameworks and libraries like angular.js, backbone.js and jQuery make development easier and time-to-market faster.

Unfortunately, there is a tradeoff for these rich web experiences. As web pages become bloated with JavaScript, there are substantial delays in application delivery performance — particularly on mobile devices where there are smaller CPUs, memory and cache sizes. It’s not uncommon for end-users to wait for seconds, staring at a blank screen, while the browser downloads and parses all this code.

A big part of the bottleneck causing these performance delays lies within the delivery of JavaScript code. When a site loads and a browser request is made, traditional web delivery approaches respond by sending all of the JavaScript code without understanding how the end users’ browsers will use it. In fact, many times, more than half of the code is never even used. Developers have tried to mitigate this challenge by turning to minification – an approach that removes unnecessary data, such as white-spaces and comments. But this approach provides only minimal benefits to web performance.

Now imagine instead, if the browser could intelligently decide what JavaScript code is actually used and download that code on-demand. While the performance benefit could be substantial, demand loading code without breaking the application would be a very challenging problem. This is exactly what a new technology called SmartSequence with JavaScript Streaming from Instart Logic does. It’s the first such innovation that I have seen that applies machine learning to gain a deep understanding of how browsers use JavaScript code to optimize its delivery and enhance performance.

By using real-time learning coupled with a cloud-client architecture, their technology can detect what JavaScript code is commonly used and deliver only the necessary portions. The company says this approach reduces the download size of a typical web application by 30-40%, resulting in dramatic performance improvements for end users. With this new method, developers can now accelerate the delivery of web applications even as the use of JavaScript continues to rise.

For web and application developers, this gives them the freedom to push the boundaries of web development without sacrificing performance, opening up endless opportunities for revolutionizing web and mobile applications. The way Instart Logic is looking to solve this problem is interesting as I haven’t seen this approach in the market before, so it’s definitely one to watch as it evolves. For more details on the technology, check out the company’s blog post entitled, “Don’t Buy the JavaScript You Don’t Use.”