基于非线性差分随机早期检测的拥塞控制机制

主动式队列管理技术是网络中间节点拥塞控制的关键技术,它和基于窗口的TCP端到端的拥塞控制相结合,是解决目前网络拥塞控制问题的有效途径。采用非线性增长数据包丢弃概率与差分服务模型相结合的方法,提出NLD_RED算法。该算法是对RED改动很小的一种算法,实现了控制参数动态化,一定程度上消除了RED的参数敏感性问题,满足了不同优先级差分服务需求,增强了调节拥塞控制的能力。在仿真平台上设计了仿真实例,对算法性能进行评估。算法能够尽量维持平均队列长度在一个相对稳定状态,提高了算法稳定性和灵活性。

Non-linear differentiate random early detection based on congestion control mechanism

Active queue management is a critical congestion control technology used by middle nodes of the networks.The windows-based TCP end-to-end congestion control mechanism is widely used by the end nodes.The combination of the two mechanisms is a efficiency way to solve the congestion control problem in networks.By combining the non-linear increase of the drop probability of the packets with the differentiate service model,a novel arithmetic,called Non-Linear Differentiate Random Early Detection,is proposed.NLD_RED is a dent from RED,which realizing the dynamic of the parameters control,eliminating the parameters sensitivity problem of the RED in some degree,meeting the different precedence levels of the users,enhancing the ability of congestion control adjustment.To evaluate the performance of the proposed arithmetic,We design a scenario on the simulation platform.The results of the simulation show that the arithmetic maintains the average queue length in a relatively stable state and improves the stability and agility.