CS2P - Improving Video Bitrate Selection and Adaptation with Data-Driven Throughput Prediction

Posted on | In | | Visitors
A paper a day keeps the doctor awake.
  • Link2Paper
  • [Author: Yi Sun et al.]
  • [Keywords: Multimedia streaming, Transport protocols, Network measurement]

Motivation

Bitrate adaptation is critical to ensure good quality-of- experience (QoE) for Internet video. Several efforts have argued that accurate throughput prediction can dramatically improve the efficiency of:

  1. initial bitrate selection to lower startup delay and offer high initial resolution
  2. midstream bitrate adaptation for high QoE.

However, prior ef- forts did not systematically quantify real-world throughput predictability or develop good prediction algorithms.

Contribution

  1. Insights from iQiYi dataset w/ 20M+ sessions:
    • Sessions sharing similar key features (e.g., ISP, region) present similar initial throughput values and dynamic patterns;
    • There is a natural “stateful” behavior in throughput variability within a given session.
  2. Based on these insights, the authors develop CS2P, a throughput prediction system which uses a data-driven approach to learn
    • clusters of similar sessions
    • an initial throughput predictor
    • a Hidden-Markov-Model (HMM) based midstream predictor modeling the stateful evolution of throughput

HMM

  1. Development of a prototype system and show using trace-driven simulation and real-world experiments:
    • CS2P outperforms existing prediction approaches by 40% and 50% in terms of the median prediction error for initial and midstream throughput
    • CS2P achieves 3.2% improvement on overall QoE and 10.9% higher aver- age bitrate over state-of-art Model Predictive Control (MPC) approach [1], which uses Harmonic Mean for throughput prediction.

Evaluation

Discussion

Limitations

  1. Proprietary dataset from iQiYi :/
  2. Off-line training and online prediction might not be general enough

Online Learning

Sensitivity Analysis

Network function models effectively reason about network configuration

A bunch of listed papers on the top right paragraph in the first page. Reference in the original paper [24, 25, 30–32, 38, 39, 46, 52, 55, 59].

NF verified written in Click [2]

Dobrescu, M., and Argyraki, K. Software Dataplane Verification. In Symp. on Networked Systems Design and Implem. (2014) [1]

Ref