Machine-learning algorithms are everywhere. They’ve permeated the way platforms work in the cloud, in analyzing Big Data, and in making AI more intelligent. Now, a software-defined application delivery startup is using machine learning to streamline code performance.
SmartSequence, the new HTML and JavaScript streaming technology from Instart Logic, uses real-time machine-learning techniques to sift through execution code in a Web application, and let the browser load only the code that’s needed. The machine-learning algorithm learns from user behavior to identify patterns in code use, streaming, and instrumenting only the relevant HTML or JavaScript code to reduce download size and accelerate website performance.
“SmartSequence is this new technology that uses a combination of machine learning and our architecture to make code on the Web perform faster,” said Peter Blum, Instart Logic vice president of product management. “So how it works is the end user types a website into the browser as they normally would, and then on the back end SmartSequence will analyze and then instrument the code, measuring based on real end users which JavaScript functions are used and in what order. The SmartSequence technology is continually learning how the JavaScript code is executed across different users.”
Instart Logic is a late-stage startup founded in 2010 that takes an application-centric view of software performance. Bolstered by funding rounds of US$9 million in 2012, $17 million in 2013 and $26 million in 2014, the company has developed a cloud-client architecture as part of its Software-Defined Application Delivery (SDAD) platform. By running a virtualization layer directly in the end user’s browser, the platform “streams” or itemizes the components of a Web page down into how each piece of HTML or JavaScript code is being used in a Web application to gain concrete data and insight on the client level.
Instart Logic’s cloud-client architecture model underlying SmartSequence.
Blum explained that as websites and applications have become more sophisticated—using JavaScript and dynamic HTML to enable more responsive design, or creating native application experiences in the browser—sites push logic down to the client-side, and browsers essentially become application execution environments handling a growth in code. According to HTTP Archive, since 2011 the average size of HTML code transfers and the number of requests have doubled, while those for JavaScript have tripled.
“There’s a big increase in use of JavaScript across the board, especially in adoption of frameworks and libraries like jQuery, Angular, Backbone, Node, Ember and others,” said Blum. “The problem with all this JavaScript and dynamic HTML code is that a lot of it is never used. People are using these great third-party libraries like jQuery that do a hundred things, but the reality is most developers are using maybe 10 of them. We’re delivering a lot of JavaScript code that’s never used, and also when developers write internal JavaScript older code is often not removed.”
(Related: ECMAScript 6 will be a milestone release for JavaScript)
Instart Logic reported that since the initial release of its SDAD platform, its customers have seen a 60% reduction in page load time, and that thus far SmartSequence has added an additional 15% boost in HTML and JavaScript execution speed. Blum believed that as the machine-learning capabilities of SmartSequence evolve to understand increasingly complex JavaScript functions—especially given the impending standardization of ECMAScript 6—the developer will have less to do.
“When we started out, we thought this would benefit Web applications and SaaS apps the most,” Blum said. “But as everyone has done this responsive Web redesign, e-commerce and even news and information sites are using a lot more JavaScript. There’s a tension with developers where they want to utilize all these new JavaScript libraries without paying the performance penalty. The point of SmartSequence is that we’re saying…go ahead. We’ll figure out the segment of those libraries that you’re actually using.”
