Practical coupled rate control with co-bottleneck for video streaming

Song, Jinhui

Permalink

https://hdl.handle.net/2142/124644 Copy

Description

Title

Practical coupled rate control with co-bottleneck for video streaming

Author(s)

Song, Jinhui

Issue Date

2024-04-08

Director of Research (if dissertation) or Advisor (if thesis)

Hu, Yih-Chun

Doctoral Committee Chair(s)

Hu, Yih-Chun

Committee Member(s)

• Godfrey, Philip Brighten
• Wang, Gang
• Mittal, Radhika

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Coupled rate control
• QoE
• shared bottleneck detection

Abstract

In recent years, video traffic becomes a large portion of the Internet traffic. Due to the growth of the number of users, concurrent video sessions with diverse bitrate demands have been more and more common. When these video streams share the same bottleneck link (co-bottleneck), current transport protocols do not perform well due to the bandwidth waste and instability caused by the flow competition from TCP congestion control, and the on-off traffic pattern due to current video streaming protocol. Current works include distributed and centralized approaches. Distributed approaches struggle to provide a good coordination among flows, due to their distributed nature, and fail to avoid the bandwidth inefficiency due to flow competition; centralized solutions require a large deployment overhead, due to the centralized control of both application and transport layers in a large scale. This thesis considers how to tackle this issue by joint rate control on the co-bottlenecked flows with least system overhead. The system is composed of three parts: FlowBot, a co-bottleneck detector, Cowfish, a flow shaping system, and Carven, an overall rate control system for video streaming. We first study how to detect the co-bottleneck link among flows efficiently and accurately (FlowBot), then study how to enforce weighted fairness to the flows without affecting the TCP friendliness to the cross traffic (Cowfish), and finally propose an integrated joint rate control system for video streaming with diverse bitrate demands (Carven). Our evaluations show that: FlowBot is more accurate, faster, and more efficient than the traditional passive cobottleneck detectors, thus more suitable for a large number of video streams from CDN; Cowfish achieves fast adaptation, steady throughput and TCP friendliness; Carven significantly improves the video quality for high-demand users while maintaining the quality for low-demand users. Therefore, we demonstrate a system that provides better and faster joint rate control for video streaming requiring the least support from the video application.

Graduation Semester

2024-05

Type of Resource

Thesis

Copyright and License Information

2024 Jinhui Song

Owning Collections