• DocumentCode
    2926075
  • Title

    Bayesian piggyback control for improving real-time communication quality

  • Author

    Xiao, Wei-Cheng ; Chen, Kuan-Ta

  • fYear
    2011
  • fDate
    10-12 May 2011
  • Firstpage
    1
  • Lastpage
    5
  • Abstract
    The critical tasks in designing a real-time communication system, such as online games, voice chat, and video conference applications, is to keep the end-to-end delay and jitter as low as possible. Delays and jitters are usually caused by network congestion and packet losses. A packet loss event will trigger the timeout retransmission mechanism at the transport layer, however, it is time consuming. Therefore, it is important to detect a loss event and do the retransmission efficiently. To achieve this goal, we propose a Bayesian loss detection mechanism based on the round trip time. We show that our loss detector can achieve high accuracy and high probability of detection as the false alarm ratio is lower than 20%. In addition, we also present a piggyback scheme for packet retransmission. Cooperated with the loss detector, the piggyback scheme can reduce the end-to-end delay and jitter to less than half of that of the original real-time communication system. Moreover, it generates few overheads to the network.
  • Keywords
    Bayes methods; data communication; real-time systems; telecommunication congestion control; Bayesian loss detection mechanism; Bayesian piggyback control; delays; end-to-end delay; jitter; network congestion; packet losses; packet retransmission; real-time communication quality; real-time communication system; timeout retransmission mechanism; Accuracy; Bayesian methods; Delay; Detectors; Histograms; Jitter; Real time systems; Bayesian inference; Online gaming; Parzen method; SCTP; VoIP;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Communications Quality and Reliability (CQR), 2011 IEEE International Workshop Technical Committee on
  • Conference_Location
    Naples, FL
  • Print_ISBN
    978-1-4577-1297-5
  • Electronic_ISBN
    978-1-4577-1296-8
  • Type

    conf

  • DOI
    10.1109/CQR.2011.5996080
  • Filename
    5996080