Global Optimization for Quality of Service in Internet Video Streaming: Design, Implementation, and Evaluation
Akshar Patel · 2024
This paper introduces and evaluates GO, a global control platform designed to enhance Quality of Service (QoS) for Internet video streaming. Unlike conventional reactive approaches, GO employs a predictive and data-driven methodology to manage a diverse range of factors such as Content Delivery Networks (CDNs), video encoding, and bitrates. By leveraging information from various ecosystem elements-such as video sessions, servers, and routers-GO aims to select optimal initial configurations and adapt quickly to network conditions. The paper details the design of GO and presents early deployment results from a video site, demonstrating its effectiveness in improving QoS metrics such as buffering ratio and average bitrate. Despite facing deployment constraints and business policy limitations, GO represents a significant advancement towards a more sophisticated global control platform for video streaming.