To Block or Not to Block: Accelerating Mobile Web Pages On-The-Fly Through JavaScript Classification Moumena Chaqfeh Muhammad Haseeb Waleed Hashmi NYUAD LUMS NYUAD
[email protected] [email protected] [email protected] Patrick Inshuti Manesha Ramesh Matteo Varvello NYUAD NYUAD Nokia Bell Labs
[email protected] [email protected] [email protected] Fareed Zaffar Lakshmi Subramanian Yasir Zaki LUMS NYU NYUAD
[email protected] [email protected] [email protected] ABSTRACT even worse for a large fraction of users around the globe who The increasing complexity of JavaScript in modern mobile live in low- and middle-income countries, where mobile ac- web pages has become a critical performance bottleneck for counts for 87% of the total broadband connections [13], and low-end mobile phone users, especially in developing regions. users solely rely on affordable low-end smartphone devices In this paper, we propose SlimWeb, a novel approach that au- to access the web [3]. On these devices, JS processing time tomatically derives lightweight versions of mobile web pages is tripled compared to desktops [36], resulting in long de- on-the-fly by eliminating the use of unnecessary JavaScript. lays [64] and a poor browsing experience [57, 70]. In contrast SlimWeb consists of a JavaScript classification service pow- with other Internet applications (such as video streaming), the ered by a supervised Machine Learning (ML) model that performance of web browsing is more sensitive to low-end provides insights into each JavaScript element embedded in hardware [34]. Nevertheless, the current status of the World a web page. SlimWeb aims to improve the web browsing Wide Web (WWW) shows a 45% increase in JS usage by a experience by predicting the class of each element, such that median mobile page, with only 7% less JS kilobytes trans- essential elements are preserved and non-essential elements ferred to mobile pages in comparison to desktop pages (475.1 are blocked by the browsers using the service.