Using Machine Learning to Optimize Web Interactions on Heterogeneous Mobile Systems

Abstract

The web has become a ubiquitous application development platform for mobile systems. Yet, web access on mobile devices remains an energy-hungry activity. Prior work in the field mainly focuses on the initial page loading stage, but fails to exploit the opportunities for energy-efficiency optimization while the user is interacting with a loaded page. This paper presents a novel approach for performing energy optimization for interactive mobile web browsing. At the heart of our approach is a set of machine learning models, which estimate at runtime the frames per second for a given user interaction input by running the computation-intensive web render engine on a specific processor core under a given clock speed. We use the learned predictive models as a utility function to quickly search for the optimal processor setting to carefully trade responsive time for reduced energy consumption. We integrate our techniques to the opensource Chromium browser and apply it to two representative mobile user events: scrolling and pinching (i.e., zoom in and out). We evaluate the developed system on the landing pages of the top-100 hottest websites and two big.LITTLE heterogeneous mobile platforms. Our extensive experiments show that the proposed approach reduces the system-wide energy consumption by over 36% on average and up to 70%. This translates to an over 17% improvement on energy-efficiency over a state-of-the-art event-based web browser scheduler, but with significantly fewer violations on the quality of service.

Camel: Smart, Adaptive Energy Optimization for Mobile Web Interactions

“…Existing works largely optimize the initial page loading phase, but as we demonstrate below, interactions have higher energy drain and thus more potential for savings. The few works [6], [11], [17] to address interactions assume a fixed response deadline for web content, but this runs the risk of degrading the overall user experience. By contrast, CAMEL minimizes energy consumption without compromising QoE, by offering "sufficiently good" performance.…”

“…• EBS: A regression-based method for adjusting the processor frequency to meet a fixed response deadline [6]; • Phase-aware: An event-phase-based power management strategy for mobile web browsing [17]; • ML-governor: A machine-learning-based CPU frequency governor for interactive web browsing [11]; • eBrowser: This strategy puts the browser process into sleep to drop some of the input user events [12]. All the above schemes require learning on the entire training dataset for each hardware architecture.…”

“…Some more recent works try to reduce the energy footprint of web interactions by limiting the processor clock speed [6], [7], [11] or dropping some interaction events [12]. However, these solutions are suboptimal as they achieve energy savings at the cost of user experience.…”

“…We demonstrate the benefits of CAMEL by integrating it into the rendering architecture of Chromium [16] and evaluating it against four event-based web schedulers [6], [11], [12], [17]. We perform an unprecedentedly large-scale evaluation involving 1,000 testing webpages, 30 users and four distinct mobile devices.…”

Optimizing Streaming Parallelism on Heterogeneous Many-Core Architectures

“…Predictive Modeling. Recent studies have shown that machine learning based predictive modeling is effective in code optimization [43], [44], performance predicting [45], [46], parallelism mapping [20], [47], [48], [49], [50], and task scheduling [51], [52], [53], [54], [55], [56]. Its great advantage is its ability to adapt to the ever-changing platforms as it has no prior assumption about their behavior.…”