Explore a conference talk that delves into Dashlet, an innovative system designed to enhance the quality of experience in short video streaming applications. Learn about the challenges posed by non-linear video presentation and user swiping behavior in popular platforms like TikTok. Discover how Dashlet leverages insights from real-world TikTok performance studies and user swipe pattern analysis to develop a novel out-of-order video chunk pre-buffering mechanism. Understand the system's approach to determining pre-buffering order and bitrate using a simple, non-machine learning-based model of user swipe statistics. Gain insights into Dashlet's performance improvements, including outperforming TikTok by 28-101% and reducing wasted downloaded video by 30%. This talk, presented at USENIX NSDI '23 by researchers from Princeton University, offers valuable knowledge for those interested in video streaming optimization and user experience enhancement in short-form video applications.
Overview
Syllabus
NSDI '23 - Dashlet: Taming Swipe Uncertainty for Robust Short Video Streaming
Taught by
USENIX