显式丢失通告算法的实现及其性能分析

Implementation and Performance Analysis of Explicit Loss Notification

TCP协议假定链路上的数据丢失率极小，所以当数据传输中出现丢包情况时，它认为是由于网络拥塞而导致的，并且相应地采取拥塞避免机制。然而在无线网络中，由于信道衰减，移动主机切换基站等原因，经常会导致数据的丢失。针对这一问题，显式丢失通告算法（ELN）要求中间路由在发现TCP数据包丢失时，发送显式丢失通告报文，通知TCP发送方数据包的丢失是基于链路原因，而不是由于网络拥塞。后者根据显式丢失通告屏蔽拥塞控制机制，从而避免了由于不必要的触发拥塞控制而导致的性能下降。论文将在实现ELN算法的基础上，根据模拟结果对ELN算法的性能加以评估。

In wired network, transmission errors are relatively rare, TCP assumes that segment losses are due to network congestion, and triggers congestion avoidance mechanisms, which essentially make the senders window much smaller. Unfortunately, this is not true for wireless network where, due to handoff and fading channels, transmission losses are more frequent. Explicit loss notification(ELN) claims that if it were possible to explicitly indicate when a particular loss is due to errors induced by the wireless link, then TCP could be modified so as to refrain from going into congestion avoidance. This paper will implements this algorithm, and evaluates it according to the simulation results.