Performance-Aware Design of Web Application Front-Ends

Abstract

The responsiveness of web applications directly affects customer satisfaction and, as a consequence, business-critical metrics like revenue and conversion rates. However, building web applications with low response times is a challenging task. The heterogeneity of browsers and client devices as well as the complexity of today's web applications lead to high development and test efforts. Measuring front-end performance requires a deep understanding of measurement tools and techniques as well as a lot of manual effort. With our approach, developers and designers can assess front-end performance for different scenarios without measuring. We use prediction models derived by a series of automated, systematic experiments to give early feedback about the expected performance. Our approach predicts the front-end performance of real-world web applications with an average error of 11% across all major browsers.