摘要
提出了一种在接收端模拟TCP行为的组播拥塞控制协议TEARM。该协议在每个接收端独立地维护拥塞窗口并模拟TCP协议来改变窗口大小,其后将窗口值转换为期望速率,反馈给发送端,其中反馈的速率为一段时间内的加权平均。此外,使用了基于代表的机制来抑制反馈,并使用了历史打折机制来提高协议的响应性。仿真表明,该协议具有良好的TCP公平性、速率平滑性、可扩展性以及响应性,适用于流媒体组播业务的传输。
TCP emulation at receivers for multicast congestion control protocol is proposed in the paper. The protocol maintains an independent congestion window at each receiver and imitates TCP to change the window size, which is then converted into the expected rate that is a weighted average during a period of time, and then the rate is fed back to the sender. Besides, the mechanism based on the representative is used to suppress the feedbacks and the history discount mechanism is introduced to improve the responsiveness of the protocol. The simulations indicate that the protocol shows good TCP-fairness, smoothness, scalahility and responsiveness, which makes it a good choice for streaming media transportation.
出处
《计算机科学》
CSCD
北大核心
2007年第7期42-46,共5页
Computer Science
基金
国家自然科学基金项目资助(90104012)
关键词
拥塞控制
组播
TCP公平性
可扩展性
Congestion control, Multicast, TCP-friendliness, Scalability