Research Topic: Evaluating the Impact of Content Delivery and Protocol Evolution on Web QoE in LEO Satellite Networks
Motivation
Low-Earth Orbit (LEO) satellite constellations, with Starlink at the forefront, are rapidly emerging as a global solution for ubiquitous broadband, promising superior latency compared to legacy GEO satellite services. While previous network-level studies have characterized Starlink's performance, a critical research gap exists at the application layer. Existing Content Delivery Networks (CDNs), which serve the majority of web traffic, are architecturally misaligned with LEO network topology, often leading to DNS/Anycast mislocalization and performance variance. Furthermore, the performance benefits of next-generation transport protocols like HTTP/3 (QUIC) remain largely unquantified in the highly dynamic and lossy LEO environment. Video streaming services like Youtube, Twitch, Netflix are also affected by Starlink's user-PoP assignments, leading to inefficient content delivery in respect to other terrestrial counterparts. This Master's thesis addresses this gap by presenting a comprehensive, multi-layered measurement and comparative analysis of web quality of experience (QoE) over Starlink. Specifically, we characterize the performance disparities across leading CDN providers, quantify the benefits of H3 versus H2, and introduce a robust framework for measuring user-centric QoE under LEO's unique network dynamics.
Objectives
- Characterize and Decompose end-to-end Latency Across Major CDN Providers (Cloudflare, Google, Akamai, Amazon, Fastly, more…) for Starlink users.
- Quantify the Performance disparity of HTTP/3 (QUIC) over HTTP/2 (TCP) over LEO networks. Additionally, evaluate the effects of Starlink reconfiguration interval.
- Evaluate the Impact of PoP Location and Caching on locally popular content (gov sites, regional news sites, etc) across varied regions with baseline performance comparison against terrestrial ISP providers
- Evaluating QoE metrics for web browsing over Starlink - for e.g. Page Load Time, First Contentful Paint, Above-the-Fold, Speed Index.
Analyze latency towards video streaming CDN edges. Evaluate video playback delay time for various streaming service providers.
Potential Research Questions
- How well do CDN providers adjust to Starlink's geoIP feed for effective CDN server selection?
- Are there performance disparities when requesting locally popular content over globally popular ones?
- Do newer transport protocols provide better QoE on LEO satellite networks, especially for the reconfiguration intervals?
- How do Starlink's different performance regimes (content-rich, remote, sparse-edge PoPs) affect web QoE?
- How well do video streaming services perform over Starlink?
Related Literature
Investigating Web Content Delivery Performance over Starlink
A Multifaceted Look at Starlink Performance
A Browser-side View of Starlink Connectivity
A Global Perspective on the Past, Present, and Future of Video Streaming over Starlink
Contact
Please send your application by email to me at rohan.bose@tum.de