WEEC: Web Energy Efficient Computing: A machine learning approach

Abstract

The growing popularity of web browsing on mobile devices running on a variety of embedded processors places heavy demands on already energy constrained systems. Energy consumed by web browsers is greatly misunderstood by developers. High energy consumption of web applications combined with limited battery resources greatly effects user experience (UX). We propose an embedded application to reduce the energy consumption and load time of a web page by sacrificing quality of service (QoS) but maintaining user experience. Our optimizations result in improved energy consumption and thereby reducing load time of web pages. We identify the key elements of web pages that consume high amounts of energy by interpreting the web request on a server and tier the web components accordingly. A machine learning based predictive modeling technique is introduced that automatically down-sample the web page for energy efficient rendering. We also work on different optimizations of web components but the best results are achieved by re-sampling images due to their greater contribution in overall web page content. The experimental results of our approach show a minimum reduction of 24.6% energy consumption of the overall system.