摘要
提出一种基于等效活动流预测的主动队列管理(AQM)机制———近似公平丢弃(AFD)机制,通过抑制行为不端流进入队列的机会,从而获得业务流之间近似的公平.和其他现有的AQM机制不同,AFD并不丢弃低于最大允许速率门限的流的包,因此保护了行为良好的流免受行为不端流的影响,进而改善了这类流的吞吐量,降低了排队时延.仿真结果证实,在TCP,UDP流共存的情况下,AFD机制的性能优于目前典型的AQM机制,接近需要维持所有流状态信息的理想情况下的性能.
Based on equivalent active flow number estimation, we propose a novel Active Queue Management（AQM） schemne called Approximate Fairness Dropping（AFD）, which is able to achieve approximate fairness by containing misbehaved flows＇ access queue opportunity. Unlike most of the existing AQM schemes, AFD does not drop the packets whose arriving rate is within the maximum admitted rate threshold, so it protects the well-behaved flows against misbehaved ones. Moreover, it improves throughput and decreases queuing delay. Our simulations demonstrate that this new technique outperforms the current typical AQM schemes and closely approximates the ＂idea/＂ case, where full state information is needed.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2006年第1期5-10,共6页
Journal of Xidian University
基金
国家自然科学基金重大研究计划面向项目资助(90104012)
关键词
拥塞控制
主动队列管理
公平性
近似公平丢弃
等效活动流
congestion control
active queue management
fairness
approximate fairnessd ropping
equivalent active flow
