SC16 Salt Lake City, UT

67. Parallel Performance-Energy Predictive Modeling of Browsers: Case Study of Servo


Authors: Rohit Zambre (University of California, Irvine)Lars Bergstrom (Mozilla)Laleh Beni (University of California, Irvine)Aparna Chandramowlishwaran (University of California, Irvine)

Abstract: Mozilla Research is developing Servo, a parallel web browser engine, to exploit the benefits of parallelism and concurrency in the web rendering pipeline. Parallelization results in improved performance for pinterest.com but not for google.com. Occasionally, the overhead of creating, deleting, and coordinating parallel work outweighs its benefits. In this poster, we showcase results of our models, generated through supervised learning, that capture the relationship between key web page primitives and a browser's parallel performance. Such a model allows us to predict the degree of parallelism in a web page. Additionally, we consider energy usage trade-offs for different levels of speedups in our automated labeling algorithm. This is critical for improving the browser's performance and minimizing its energy usage. Experiments on a quad-core Intel Ivy Bridge (i7-3615QM) laptop, with 535 pages on Servo's layout stage, show performance and energy improvements of up to 94.52% and 46.32% respectively.

Poster: pdf
Two-page extended abstract: pdf


Poster Index