Comparing In-Browser Methods of Measuring Resource Load Times

Eric Gavaletz
Jasleen Kaur
W3C Workshop on Web Performance 8, W3C, W3C/MIT 32 Vassar Street Room 32-G515 Cambridge, MA 02139 USA (2012)

Abstract

When looking for an excellent platform for conducting end­-to-­end network performance measurement that is large­-scale and representative, researchers should look no further than the browser­­­ -- after all, browsers are installed everywhere and are used multiple times per day by most Internet users. In this work, we investigate the use of the DOM, XHR and Navigation Timing API for measuring HTTP response times within browsers, with the goal of estimating path latency and throughput. The response times are measured using a set of popular browsers in a controlled environment­­­this helps us isolate the differences between the browsers as well as study how closely the measurements match the ground truth. We show that, in general, the XHR method yields the most consistent measurements across browsers, but that the new Navigation Timing and the proposed Resource Timing APIs could change that. We also use the measurements from our controlled environment to study the impact of each of our investigated measurement methods on a hypothetical measurement study.