主动队列管理中的智能分组丢弃新机制

An Intelligent Packet Dropping Mechanism for Active Queue Management

主动队列管理通过网络中间节点有控制的分组丢弃实现了较低的排队延时和较高的有效吞吐量,是TCP端到端拥塞控制近来研究的一个技术热点.已有的大多数算法在判定分组丢弃时大都沿袭了RED的概率丢弃机制,具有一定计算复杂度的随机数生成过程不利于路由器性能的优化.在本文中,我们首先定义了拥塞指数这一新的测度变量来量化描述网络的拥塞状态.接着,利用模糊逻辑设计了一种新的智能分组丢弃机制,离线的合成推理使得分组丢弃的判定仅需要简单的查表操作和比较运算即可完成,为优化路由器的性能提供了便利.数字仿真的结果表明:智能分组丢弃机制的性能优于经典的RED算法,控制队列的能力强,鲁棒性好,稳定工作域大,能很好地抵抗突发性和非弹性业务的干扰,适合工作在瞬息万变的动态网络环境中.

Active Queue Management maintains the smaller queuing delay and higher throughput by purposefully dropping packets at the intermediate nodes, and is becoming a focus in field of TCP end-to-end congestion control. Most of the existing AQM algorithms follow the probability-based packet dropping mechanism originated from RED, which is disadvantage to the router performance optimization because random number generation needs rather computational effort.In this paper,the congestion index is put forward to quantificationally describe the network congestion state. Then, a novel intelligent packet dropping mechanism is designed based on fuzzy logic.The off-line inference greatly simplifies on-line process,and thus it is very beneficial to optimize the router performance. The simulation results show that the intelligent mechanism is superior to the classical RED algorithm. It is insensitive to load variance, and able to resist against the disturbance caused by burst and short-lived http sessions,and non-elastic UDP connections, which is very suitable for the rapidly changeful network environment.