主动队列管理算法的分类器实现

作为端到端拥塞控制的增强机制,主动队列管理(AQM)通过在网络中间节点有目的地丢弃分组来维持较小的队列长度和较高的链路利用率.已有的大多数主动队列管理算法沿用了随机早期探测(RED)算法首创的概率丢弃机制.本质上,判决是否丢弃分组的过程是一个依赖于网络拥塞状态的决策过程,因此,概率决策不应该是唯一的方法.在本文的研究中,我们首先归纳了理想AQM算法所应具备的品质,然后应用模式识别中分类器的设计思想提出了一种新颖简洁的主动队列管理策略实现框架,并基于Fisher线性判别方法为AQM设计了一个两维两类分类器(TCC).仿真试验表明TCC有效、敏捷、鲁棒,扩展性好,同时实现简单,计算开销小,有利于高速路由器的性能优化.

A Classifier Implementation for Active Queue Management Algorithm

Active Queue Management (AQM) is an enhancement mechanism to end-to-end congestion control,which can maintain smaller queuing delay and higher throughput by purposefully dropping the packets at intermediate nodes.Almost all the existing AQM schemes follow the probability dropping mechanism originated from Random Early Detection (RED).It is essentially a decision process with aid of information about network congestion that AQM decides whether or not to drop an incoming packet,thus the probability discrimination should not be only one way.We firstly summarize the properties of an ideal AQM scheme,and then develop a novel framework for AQM scheme based on the two-category classifier,which is considerably simple and extensible.A two-dimensional two-category classifier (TCC) for AQM is designed using the Fisher Linear Discriminate approach.The simulation results show that TCC outperforms other well-known AQM schemes in the integrated performance,namely TCC is an effective,prompt and robust algorithm.Additionally,TCC scheme requires few CPU cycles,which makes it deploy in high-speed routers simply.