通信学报

• 学术论文 • 上一篇    下一篇

适合TCP的网络编码重传机制

陈 静,宋学鹏,刘 芳   

  1. 1. 科学技术部信息中心,北京 100862;2. 中国科学院 计算机网络信息中心,北京 100190
  • 出版日期:2014-08-25 发布日期:2014-08-15
  • 基金资助:
    国防科技预研基金资助项目;中国科学院新型网络协议体系研究基金资助项目

Effective retransmission in network coding for TCP

  • Online:2014-08-25 Published:2014-08-15

摘要: 通过理论分析,看出基于反馈的重传方法比定量重传的方法有更低的解码延迟。提出了一种新型的基于反馈的网络编码(FNC)重传机制,利用seen机制中的隐含信息来获取接收方解码所需的重传分组个数,并改变了编码规则使部分分组可以提前解码。该机制不仅可以处理有固定误码率的随机分组丢失,还可以有效地应对大量突发性分组丢失。仿真结果显示,该机制在高误码率下也能保持较高的吞吐量,且极大地减少了解码延迟,传输过程基本不受分组丢失的影响,有效地对拥塞控制协议隐藏了链路错误。算法简单有效,更适于在实际系统中应用。

Abstract: Theoretical analysis for evaluating the decoding delay and redundancy of the different retransmission schemes show that the feedback based retransmission scheme has a lower decoding delay than the scheme with a fixed retransmission rate, whereas these schemes which retransmit packets containing all packets in the coding window are deeply influenced by the loss rate. Therefore a novel network coding retransmission scheme which uses the implicit information of seen scheme to acquire the number of packets the receiver needs for decoding packets is proposed, the encoding rules of retransmission to let a part of packets decodable in advance are changed. Proposed scheme can work well on handling not only random losses but bursty losses in reducing decoding delay. More important, it isn’t influenced by the loss rate, and thus can effectively mask losses. Simulation results show that the new scheme significantly outperforms the previous coding approach in reducing decoding delay and increasing throughput at the same time. It is more possible to be implemented in practical systems.

No Suggested Reading articles found!