Accelerating Web Start-up with Resource Preloading

Ji Hwan Yeo, Jae-Hyeon Rim, Chang Hyun Shin, Soo-Mook Moon

Abstract: Long start-up time of web apps or web pages at the client device may affect the user experience negatively. One bottleneck in reducing the start-up time is resource loading overhead. To reduce the overhead, resource preloading has been proposed, which load resources ahead of time, instead of loading them on time when they are used. For commercial client device such as smart TV, it is reasonable for the browser of the client to do resource preloading. Existing client-only technique remembers the resources accessed in the previous visit and then preloads them in the current visit. However, it preloads the resources in some arbitrary order, thus not dealing well when the preloading order is important, especially for user experience. One solution is employing a resource dependence graph, often used for preloading with proxy servers or web servers, but in client-only environment we need to compute the graph while preloading is in progress, and update the graph incrementally when the resources change. This paper proposes such a dependence graph-based, client-only resource preloading technique. For better user experience, we decide the preloading order based on those factors that affect user perception such as the size or the location of images in the screen. Our experimental results on Chromium browser with real web apps on an embedded board and on a commercial smart TV show that the proposed technique can improve the UX-related app start-up time (above-the-fold time or speed index) tangibly, allowing the user to really feel the difference.