Abstract
Video streaming dominates traffic on the global Internet. User Quality of Experience (QoE) is managed by Dynamic Adaptive Streaming over HTTP (DASH) players that request and retrieve bitrates of varying quality according to application-perceived network conditions. Selection of higher bitrate encodings is preferable, but must be balanced against other QoE indicators such as avoiding rebuffering, in effect using these indicators to create a feedback control loop. However, HTTP relies on TCP or, more recently, QUIC as transport protocols with congestion control mechanisms built around their own feedback control loops. As a result, tension emerges between application and transport feedback control loops that adversely affects multiple users when behind a shared bottleneck. Existing solutions focus on either client- and/or server-side adaptation mechanisms, or introduce additional elements into the network to orchestrate the interactions more effectively. While such solutions have proven effective, this thesis posits that they only address the symptoms of the underlying problem – nested feedback control loops.This thesis takes a holistic view that bridges separate feedback controls in the application and transport. First, an evaluation of TCP’s slow start over real user speed test data demonstrates that the congestion avoidance phase is unavoidable when sending video segments. Next, a mechanism is designed and implemented enabling the client application to signal the server transport. Evaluations show improved stability among on-demand video clients behind a shared bottleneck. Additionally, for the live streaming space, a signalling mechanism is designed from the server transport to the client application. Experimental evidence shows improved stability when contention is low and a reduction in QoE loss when contention is high. Finally, the server-side signalling is extended to QUIC to demonstrate the approach as transport agnostic. Overall, nested feedback control loops are unavoidable, but tensions may be resolved by cross-layer signalling between clients and server.
Date of Award | 3 Dec 2024 |
---|---|
Original language | English |
Awarding Institution |
|
Supervisor | Marwan Fayed (Supervisor) & Saleem Bhatti (Supervisor) |
Keywords
- Computer science
- Quality of experience
- Dynamic adaptive streaming over HTTP (DASH)
- Computer networking
- Transmission control protocol (TCP)
- QUIC
- Congestion control
- Streaming media
Access Status
- Full text open