移动网络中重传超时问题的研究

Research on Retransmission Timeout over Mobile Data Networks

作为广泛使用的网络传输控制协议,TCP(Transmission Control Protocol)在高速移动网络中遇到了新的性能瓶颈。首先由于移动网络中存在随机位错误导致的丢包,而TCP协议不能有效区分这类丢包与拥塞丢包,导致TCP频繁的降低拥塞窗口无法有效利用移动网络的带宽资源。其次,高速移动网络的发展使得带宽时延积BDP(Bandwidth-Delay Product)进一步增大,在发生丢包时TCP协议中的流量控制将导致性能瓶颈和易引起重传超时。通过Wireshark工具抓取大量的tracing进行分析,发现重传超时的主要原因是重传数据包再次被丢,而TCP又不能发现丢失原因,因此无法进行再次重传最终导致重传超时。针对这一问题,本文提出的方法 DTOR(Detect Timeout and Retransmission)可以帮助TCP检测到重传数据包再次丢失并触发再次重传,DTOR使网络带宽利用率提升了20%左右。

Recent advances in high-speed mobile networks have revealed new bottlenecks in ubiquitous TCP protocol deployed in the Internet. First, due to the existence of random bit errors in the mobile network, and TCP protocol can’t effectively distinguish non-congestive loss from congestive loss, resulting in TCP frequently reduce the congestion window, and can’t effectively use the mobile network bandwidth resources. Second, the development of high-speed mobile network makes the bandwidth delay BDP （Bandwidth-Delay Product） to further increase, when the packet loss occurs, TCP protocol flow control will also lead to performance bottlenecks and retransmission timeout. Using the Wireshark tool to capture a large number of tracing for analysis, found the main reason for the retransmission timeout is that the retransmission packet is lost again, and TCP sender can’t find the cause to the loss, so loss packet can’t be retransmitted again by TCP sender, eventually leading to RTO. In response to this problem, Optimization techniques – DTOR （Detect Timeout and Retransmission） can help TCP detect that the retransmitted packet is loss again and triggers TCP sender retransmission again. Using emulated experiments showed that the proposed optimization techniques sufficiently utilize the bandwidth.