摘要
提出一种基于自相似长相关性网络流量的TCP拥塞控制机制。利用自相似长相关性网络流量极值出现时间长的特性,预测下一时段的网络流量,动态调整相应的拥塞窗口a值,使TCP拥塞窗口从简单的线性增长变成随着网络资源的具体情况呈阶梯函数改变。仿真模拟表明,该方案降低了多媒体业务流的突发性,平滑了数据速率的变化,使缓存队列长度下降,增加了总体数据的传输率,提高了带宽的利用率。控制网络拥塞,减小延迟,达到网络负载平衡。
A type of TCP congestion control mechanism is presented based on self-similar and Long Range Dependence （LRD） of network traffic.Next period of network traffic has been forecasted through the use of self-similar and LRD that the appearance of the extremum lasts long in network traffic.The congestion window dynamically adjusts the value of a,which makes TCP congestion window change from simple linearity to trapezoidal function followed the network resource.The simulations show that this scheme can reduce the outburst of the multimedia operation flow,and therefore make the speed of data smoothly change.Consequently,the length of buffer queue declines,and the rate of data transfer is increased to improve the usage efficiency of bandwidth.As a result,the scheme may control network congestion and reduce delay to gain the balance of the network load.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第21期92-94,共3页
Computer Engineering and Applications
基金
国家863高技术研究发展计划资助项目(编号:2003AA712032)
关键词
拥塞窗口
流量预测
自相似
缓存队列
RTT
congestion window, traffic forecasting, self-similar, buffer queue, RTT
作者简介
薛建生（1954-）,女,东北大学博士研究生,辽宁大学教授,主要研究方向：计算机网络.E-mail：xuejs@lnu.edu.cn
王光兴（1937-）,男,东北大学教授,博士生导师.
