Blog Header Image: Learning Engine

Learning Engine Helps You Learn From Real Users to Make Your Website Faster

November 14, 2017 · by Ted Shuter ·

In an earlier blog post, I gave an overview of Adaptive Acceleration, Akamai's intelligent solution to continuously apply performance optimizations with zero development effort.

Now I want to dive into one particular feature of Adaptive Acceleration that I think you’re going to love. It’s called Learning Engine, and you can probably guess by the name what it does.

Learning Engine learns from real users to continuously and automatically optimize performance. In particular, Learning Engine:

  • Uses user navigation and resource timing data to determine optimizations
  • Adapts to changes in browsing behavior based on the device, network, and browser
  • Continuously updates to adapt optimizations to changes
  • Offers simple one-step configuration to enable across all Akamai properties

How does Learning Engine work?

Learning Engine uses real-user data to automatically apply optimizations. Using real-user monitoring (RUM) data from Akamai mPulse, the system uses both Navigation Timing and Resource Timing beacons to understand all of the resources, when they were needed, and by which user-agent.

This approach continuously looks at your site to adapt to changes in the site or user behavior. Using RUM data also circumvents issues that are common with other analysis techniques and makes sure that the analysis sees the same resources that users are seeing in the wild. One common example is analysis of sites that are only available for logged-in users.

What's new in the latest version of Learning Engine?

With the most recent release, you now have the ability to understand the critical path, and separate analysis by user-agent:

  • By understanding the critical path resources, the engine can focus on the key resources needed to render faster by excluding things that are not critical. For example, Learning Engine can automatically ignore print style sheets.
  • The second area is analysis by user-agent; this allows Learning Engine to look at how a site behaves for mobile and desktop variants of the same site to optimize server-side response pages.

Ultimately, this means different optimizations can be applied for these different users.

Overall, Learning Engine is the heart and soul of Adaptive Acceleration. It’s a radical new approach to performance optimization that enables the system to understand how a site is composed, largely replicating what a performance engineer would do looking at a waterfall.

For additional insight and information on Learning Engine and Adaptive Acceleration, see our video of the Developer General Session presentation at Edge 2017 (the Adaptive Acceleration portion begins at 33:18).

Adaptive Acceleration is available with Ion, Ion Premier, and Ion Media Advanced. To get started with Adaptive Acceleration right now just log into your Akamai customer portal, where you can try Ion via the Marketplace. Or if you’re an existing Ion customer, simply activate Adaptive Acceleration by flipping the switches to “On”.

See the other posts in this four-post series about Adaptive Acceleration:

  1. Overview
  2. Learning Engine (you are here)
  3. Resource Optimizer
  4. Script Management