ISSN 1000-1239 CN 11-1777/TP

• 论文 • 上一篇    下一篇

基于TCP NewReno的稳态吞吐量分析模型

孙伟1 温涛2 冯自勤3 郭权2   

  1. 1(东北大学软件中心 沈阳 110004) 2(东北大学东软信息学院 辽宁大连 116023) 3(匹兹堡大学数学系 美国宾西法尼亚洲匹兹堡市 15260) (sunwei@neusoft.edu.cn)
  • 出版日期: 2010-03-15

Steady State Throughput Modeling of TCP NewReno

Sun Wei1, Wen Tao2, Feng Ziqin3, and Guo Quan2   

  1. 1(Software Center, Northeastern University, Shenyang 110004) 2(Neusoft Institute of Information, Northeastern University, Dalian, Liaoning 116023) 3(Department of Mathematics, University of Pittsburgh, Pittsburgh PA 15260, USA)
  • Online: 2010-03-15

摘要: 最近的研究表明,在当前网络未启用SACK选项的TCP流中,有超过一半的数据流采用TCP NewReno的快速恢复算法.而Padhye提出的基于TCP Reno的TCP吞吐量分析模型,不能准确反映TCP NewReno数据流的吞吐量.Padhye模型在建立过程中采用猝发性丢包模型,同时忽略了快速恢复阶段和超时后的慢启动阶段,影响了预测的准确性.基于此,提出了一种基于TCP NewReno 的吞吐量分析模型.在分析过程中,采用了更符合真实网络丢包特征的丢包模型,并且充分考虑了快速恢复阶段和超时后的慢启动阶段对吞吐量的影响.仿真实验表明,该模型可以准确地预测TCP NewReno数据流的吞吐量.

关键词: TCPNewReno, 拥塞控制, 吞吐量模型, 快速恢复, 网络协议

Abstract: The authors present a simple and accurate stochastic model for the steady-state throughput of the slow but steady variant of TCP NewReno. The model describes the relations between TCP NewReno throughput and round trip time, rate of packet loss and retransmission timeout value. This work motivated by the recent measurement studies indicates that there are more than half of TCP streams which use TCP NewReno instead of Reno fast recovery with a non SACK receiver. The proposed model builds upon the approach proposed by Padhye for TCP Reno throughput modeling but extends it by capturing the effect of the fast recovery algorithm and taking into consideration the slow start phase after timeout expiration. The measurement studies suggest that these behaviors are important from a modeling perspective. And the proposed model is formulated by using a new loss model instead of foregoing loss model used by Padhye, which can better represent the loss scenarios encountered by TCP on the Internet. Validation by NS2 simulator shows that using Padhyes model to estimate TCP NewReno throughput may introduce significant error while the proposed model is able to accurately predict the steady-state throughput for TCP NewReno over a wide range of network conditions.

Key words: TCP NewReno, congestion control, throughput model, fast recovery, network protocol