6th International ICST Conference on Collaborative Computing: Networking, Applications, Worksharing

Research Article

Agiler: A P2P live streaming system with low playback lag

Download665 downloads
  • @INPROCEEDINGS{10.4108/icst.collaboratecom.2010.13,
        author={Dongbo Huang and Jin Zhao and Xin Wang},
        title={Agiler: A P2P live streaming system with low playback lag},
        proceedings={6th International ICST Conference on Collaborative Computing: Networking, Applications, Worksharing},
        publisher={IEEE},
        proceedings_a={COLLABORATECOM},
        year={2011},
        month={5},
        keywords={Educational institutions Monitoring Servers},
        doi={10.4108/icst.collaboratecom.2010.13}
    }
    
  • Dongbo Huang
    Jin Zhao
    Xin Wang
    Year: 2011
    Agiler: A P2P live streaming system with low playback lag
    COLLABORATECOM
    ICST
    DOI: 10.4108/icst.collaboratecom.2010.13
Dongbo Huang1,*, Jin Zhao1,*, Xin Wang1,*
  • 1: School of Computer Science, Fudan University, China, Shanghai Key Lab of Intelligent Information Processing, Shanghai, China
*Contact email: 082024082@fudan.edu.cn, jzhao@fudan.edu.cn, xinw@fudan.edu.cn

Abstract

Short playback lag is preferred in many urgent and interactive scenarios such as live sports and distance education. However, measurement studies have shown that many popular P2P live streaming systems still suffer from long playback lag, say, more than 100 seconds, which makes the live streaming less realtime. Due to the unstructured nature of P2P networks, it is really a challenge to reduce the playback lag in P2P live streaming system under limited server bandwidth. In this paper, we propose a novel P2P live streaming system - Agiler that aims at reducing the playback lag in lag-sensitive applications under limited server bandwidth. In Agiler, we first group the peers into clusters according to the Autonomous System Number (ASN) and then spread the broadcast in ripples that gradually increase in playback lag as we move away from the media server. It adopts synchronous playback within a cluster to accelerate the chunk diffusion and asynchronous playback among different clusters to facilitate the chunk swarming. Combined with the partitioned buffer strategy and the proportional playback lag strategy, the newly generated fresh chunks can be delivered to all peers in time. The proposed system is evaluated through extensive packet-level simulations, which show the bandwidth utilization ratio of the peers is improved and the average playback lag is significantly reduced.