一种基于柯西分布的网络拥塞控制策略  

A Network Congestion Control Strategy Based on the Cauchy Distribution

作　　者：雷岚栋 池亚平[2] 张亮亮 LEI Landong;CHI Yaping;ZHANG Liangliang(Xidia n University,Xi'an 710071,P.R.China;Beijing Electronic Science and Technology Institute,Beijing 100070,P.R.China)

机构地区：[1]西安电子科技大学,西安市710071 [2]北京电子科技学院,北京市100070

出　　处：《北京电子科技学院学报》2021年第2期32-42,共11页Journal of Beijing Electronic Science And Technology Institute

基　　金：国家重点研发计划项目(课题编号:018YFE0200600)资助。

摘　　要：当某些网络资源的网络负载超出其容量时,通信网络可能会发生拥塞,从而导致拥塞的网络设备丢弃数据包,拥塞控制机制在保持网络稳定性和实现网络资源的高利用率方面起着重要作用。在高带宽延迟网络中,TCP所使用的AIMD算法窗口抖动剧烈,导致其稳定性弱,系统吞吐量小,拥塞发生时,拥塞窗口剧烈减小。为此,本文提出一种基于柯西分布的网络拥塞控制策略,该策略主要是根据采样RTT数值的统计特性,迭代统计模型之柯西分布函数,动态更新预测下一次的拥塞窗口,使网络能在较短的时间达到较高的吞吐量;通过NS-3仿真对所提出的策略进行了验证。结果表明,与TCP New Reno相比,此策略有效提高了网络稳定性和网络性能。When the network load of a network resource exceeds the capacity,congestion may occur inthe communication network,causing equipment in the congested network to drop packets.Congestioncontrol mechanism plays an important role in maintaining network stability and realizing high utilizationrate of network resources.In a network with high bandwidth delay,AIMD algorithm used by the TCPhas a severe window jitter,leading to poor stability,low sysltem throughput,and shapely reduced congestion window when the congestion occurring.To address these problems,a network congestion controlstrategy based on thc Cauchy distribution is proposed in this paper.The strategy is mainmly hased on the statistical properties of sampling RIT values and utilizes the Cauchy distribution function in iterative statistical model Next congestion window is forecasted and updated dynamically to help the network reach a high throughput in a short time Validity of the proposed strategy is verified by NS3 simulations and simulation results show that the strategy is indeed effective in improving network stability and increasing system throughput compared with the TCP New Reno.

关 键 词：拥塞控制 预测统计模型 柯西分布 NS3仿真 

分 类 号：TP393.06[自动化与计算机技术—计算机应用技术]