TCP-Rab的吞吐量模型及实验研究  

Modeling of TCP-Rab throughput and its empirical study

作　　者：汤旭红[1] 刘正蓝[2] 朱淼良[2] 

机构地区：[1]上海海运学院航海系,上海200135 [2]浙江大学计算机系,浙江杭州310027

出　　处：《通信学报》2004年第8期24-35,共12页Journal on Communications

基　　金：上海市科技发展基金资助项目(022112186)

摘　　要：基于接收端通告的TCP(TCP-Rab,receiver advertisement based TCP)协议是我们实现的一种新的TCP协议,在文献[1]中对TCP-Rab的算法设计、实现进行了详细阐述。本文对TCP-Rab进行了少量改进,重点导出了TCP-Rab的吞吐量性能模型,并对TCP-Rab进行了试验研究。该模型采用统计的方法,在导出TCP-Rab的吞吐量性能模型的时候,不仅考虑了TCP连接的拥塞避免阶段对吞吐量的影响,也考虑了慢启动阶段对吞吐量的影响,同时还考虑了一个发送窗口内多个数据包随机丢失对吞吐量的影响,因此该模型能适用于实际的网络环境中。In our previous work, a new TCP protocol which is named receiver advertisement based TCP (TCP-Rab) is proposed. The design and implementation of Rab's algorithm are described in details in the literature [1]. In this paper, we made only a little modification on TCP-Rab, the pivot issue of this paper is focused on the throughput model and empirical study of Rab. In the procedure of the deduction, we adopt a statistical technique. This model takes account of not only the effect of congestion avoidance phase, but also the effect of slow start phase on TCP throughput, and also the effect of multiple random packet loss in one window, so this model is competent for actual Internet.

关 键 词：TCP协议 吞吐量模型 带宽估计 立即恢复 

分 类 号：TP393.04[自动化与计算机技术—计算机应用技术]