Talk
Talk by Sapna Chaudhary
Title: Measurement, Analysis, and Intelligent Optimization of Streaming Media QoE over Modern Browsers
Abstract:
Video streaming constitutes the majority of today's Internet traffic, making efficient transport protocols essential for delivering high Quality of Experience (QoE) across diverse network environments. While QUIC and HTTP/3 have been widely adopted to improve web performance, their real-world behavior depends not only on protocol design but also on browser implementations and varying network conditions. This work investigates the performance of QUIC-enabled video streaming through large-scale Internet measurements and develops intelligent mechanisms to improve streaming QoE. To achieve this, it introduces H3B, an HTTP/3-compatible browser-based measurement framework and dataset for analyzing QUIC behavior across modern browsers and heterogeneous networks. The study identifies key browser implementation choices that significantly influence streaming performance and proposes browser-level optimizations to mitigate these limitations. Building on these insights, this work presents IntelSwitch, a reinforcement learning-based framework that dynamically selects between QUIC and TCP based on prevailing network conditions by integrating protocol-aware decision-making into the Chromium browser and the dash.js video player. Extensive evaluations demonstrate significant QoE improvements across both bandwidth-constrained and high-speed networks. Overall, this work combines Internet measurement, browser systems, and machine learning to advance adaptive transport protocol selection and improve multimedia streaming performance in next-generation network environments.
Speaker Bio:Sapna Chaudhary recently completed her Ph.D. in Computer Science and Engineering from IIIT Delhi. Her research focuses on computer networks, Internet measurement, next-generation transport protocols, and adaptive video streaming systems. She investigates the performance of transport protocols across diverse network conditions and has developed reinforcement learning-based frameworks for intelligent transport protocol selection to improve multimedia Quality of Experience (QoE). Her work combines large-scale experimental evaluation with system development, including extensive modifications to the Chromium browser and DASH.js player. She has published in venues such as IEEE Transactions on Network and Service Management (TNSM), NeurIPS (Datasets and Benchmarks), and COMSNETS. Her research has been recognized through competitive travel awards from Microsoft Research, N2Women, SIGCOMM, ARCS, COMSNETS, IMC, MobiSys, and MMSys.
- 10 - 10 July,2026
- 11:00 am - 12:00 pm
- Bharti 501