基于RED算法的非线性拥塞控制

由于RED算法是采用丢包率随平均队列长度线性变化的方法，因此导致网络在拥塞并不严重的时候丢包率较大，在拥塞比较严重的时候丢包率较小，拥塞控制能力较低。该文提出非线性平滑算法通过对RED算法的丢包率函数进行非线性平滑，在最小阈值时丢包率增长速度比较小，在最大阈值时丢包率增长速度比较大，有效地控制了平均队列长度，具有较好的拥塞控制能力。NS2仿真结果表明该算法对丢包率、端到端时延、吞吐量以及时延抖动等性能均有较明显的提高。

Non-linear Congestion Control Based on RED Arithmetic

Because RED adopts the way that drop packet ratio varies following average queue length, as a result, drop packet ratio is high when the network congestion is not serious and drop packet ratio is low when the network congestion is serious, so the congestion ability is not effective. This paper puts forward a non-linear congestion control arithmetic for this shortcoming, and the arithmetic can control average queue length effectively by non-linear control to the drop packet function. The non-linear RED arithmetic has made a visible improvement on drop packet ratio, time delay, througput, time delay jitter by NS2 simulation, and result proves the arithmetic is effective.