Elsevier

Computer Communications

Volume 30, Issue 18, 10 December 2007, Pages 3778-3786
Computer Communications

Dynamics analysis of TCP Veno with RED

https://doi.org/10.1016/j.comcom.2007.09.004Get rights and content

Abstract

This paper studies the dynamics of TCP Veno with the queue management of RED (Random Early Detection). We develop a fluid-flow model to describe the behaviors of TCP Veno in wired/wireless networks. This model is further linearized to study TCP Veno’s stability issue through the linear feedback control theory. The analysis points out how the RED queue oscillates under different network parameters such as link capacity, round-trip time. Simulations are carried to validate our theoretical analysis. Furthermore, based on the analysis results obtained in this paper, we are able to provide guidelines for tuning RED parameters to stabilize the router queue, and improve the co-existence between TCP Veno and TFRC (TCP-friendly Rate Control) flows.

Introduction

TCP is a connection-oriented, reliable and in-order transport protocol. It is known that the current legacy TCP, namely TCP Reno [1], suffers from performance degradation in the wireless networks due to the lack of differentiation between the random and the congestion losses. TCP Veno [2], a sender-side TCP enhancement, was proposed to solve this problem and adopted by Linux Kernel since version 2.6.18 [3]. Veno has the ability to identify network states and adjust the additive increase multiplicative decrease (AIMD) strategy to tackle random losses. Specifically, Veno estimates the number of backlogged packets, N, at a router byN=cwndBaseRTT-cwndRTT×BaseRTTwhere cwnd is the TCP congestion window. BaseRTT is the minimum of measured round-trip time and reset when packet loss is detected. RTT is the actual round-trip time of a tagged packet. Veno compares N with a threshold parameter, β to identify network states. If N  β, the network is said to have evolved into a congestive state, and packet loss occurring in this state is considered as a congestion loss. Otherwise, the network is in a non-congestive state, and packet loss occurring in this state is considered as a random loss. In the current Veno implementation, the parameter β is set to 3. Veno adjusts its AIMD algorithm based on the network states,

Multiplicative decrease algorithm:

  • If (N < β)  cwnd = cwnd · 4/5

    • else  cwnd = cwnd/2

Additive increase algorithm:

  • If (N < β)  cwnd + = 1/cwnd for every new ACK received

    • else  cwnd + = 1/cwnd for every other new ACK received

For the random loss, TCP Veno reduces the cwnd by a smaller amount (4/5). Ref. [2] shows that, any factor greater than 1/2 but smaller than 1 can be used, so that the cutback in window size is less drastic than the case when the loss is due to congestion. Experiment results prove that the factor of 4/5 is a good setting. More mechanism and experiment details can be seen in [2].

In this paper, we study the dynamics of TCP Veno with the queue management of RED (Random Early Detection) [5]. We make use of the modeling techniques in papers [9], [10], [12], [13] to develop a fluid-flow model of TCP Veno with RED in wired/wireless networks. This model is further linearized around the equilibrium point to study TCP Veno’s stability issue. By applying the classical linear feedback control theory, the analytical results reveal how the RED queue oscillation is affected by different network parameters such as link capacity, round-trip time. Simulations are carried to validate our theoretical analysis. Furthermore, based on the analysis results obtained in this paper, we are able to provide guidelines for tuning RED parameters to stabilize the router queue, and improve the co-existence between TCP Veno and TFRC (TCP-friendly Rate Control) flows.

The rest of this paper is organized as follows. In Section 2, we review the fluid-flow model in papers [9], [10], [12], [13] firstly, and then derive the model of TCP Veno. In Section 3, we convert the fluid-flow model to a linear feedback system and apply the linear feedback control theory to analyze its stability issue. In Section 4, simulations are carried to validate our analysis and show how to tune the RED parameters for improving the co-existence of TCP Veno with TFRC. Section 5 concludes this paper and discusses future work.

Section snippets

Fluid-flow model of TCP Veno with RED

In this section, we develop the fluid-flow model of TCP Veno with RED in heterogeneous networks. Our model is composed of two parts. The first part: network and RED queue management model is the same as [12], [13]. We start with a brief description of them, and then in the second part we model the behavior of TCP Veno’s window evolution.

Theoretic analysis of the fluid-flow model

In this section, we use the linear feedback control theory to analyze the fluid-flow model derived in Section 2. Since this model is nonlinear in nature, we need convert it to a linear model, and then study its stability issue.

Simulation and validation

In this section, we use network simulator NS-2 [17] to validate our analysis, and furthermore, show how to use the theoretical results to guide the co-existence of TCP Veno with TFRC flows. TFRC [11] is a rate-based congestion control mechanism. It continuously measures the round-trip time and packet loss rate, and then uses TCP response function to calculate its sending rate. The TCP response function is expressed as follows [11],T=sR2p3+tRTO33p8p(1+32p2)where T is the transmission rate in

Conclusions and future work

This paper studies the dynamics of TCP Veno with RED queue management. A fluid-flow model is developed to describe the behaviors of TCP Veno in wired/wireless networks. The model is further linearized to study TCP Veno’s stability issue. The subsequent analysis points out how the RED queue oscillation is closely related to network link capacity, round-trip time and other network parameters. The simulation results agree with the theoretical analysis. In the future, we are considering to use this

Ke Zhang received the B.Eng. degree in Computer Science from the University of Science and Technology of China (USTC), Hefei, PR China, in 2003. Now, he is a Ph.D. student in Nanyang Technological University (NTU), Singapore. His supervisor is Dr. Cheng Peng Fu, an Assistant Professor in NTU. His research interests include congestion control and streaming in the next generation communication networks.

References (11)

  • V. Jacobson, Congestion avoidance and control, in: Proc. SIGCOMM 1988, Stanford, CA, (1988)...
  • C.P. Fu et al.

    TCP Veno: TCP enhancement for transmission over wireless access networks

    IEEE J. Select. Areas Commun.

    (2003)
  • ...
  • S. Floyd et al.

    Random early detection gateways for congestion avoidance

    IEEE/ACM Trans. Networking

    (1993)
  • C.V. Hollot, V. Misra, D. Towsley, Wei-Bo. Gong, A control theoretic analysis of RED, in: Proc. IEEE Infocom,...
There are more references available in the full text version of this article.

Cited by (6)

Ke Zhang received the B.Eng. degree in Computer Science from the University of Science and Technology of China (USTC), Hefei, PR China, in 2003. Now, he is a Ph.D. student in Nanyang Technological University (NTU), Singapore. His supervisor is Dr. Cheng Peng Fu, an Assistant Professor in NTU. His research interests include congestion control and streaming in the next generation communication networks.

Cheng Peng Fu received the B.Eng. and M.Phil. degrees in electromagnetic theory and microwave technology, and the Ph.D. degree in information engineering from the Shanghai University of Science and Technology, Shanghai, China, and the Chinese University of Hong Kong, Hong Kong, in 1990, 1995, and 2001, respectively.

He is one of founding members of CERNET (China Education and Research Network), which was launched in 1995 to network 1000 universities in China by TCP/IP technology while he served Shanghai Jiao Tong University as a Faculty Member at Networking Center from 1995 to 1997. Meanwhile, he also designed and deployed Shanghai Jiao Tong Campus ATM Network, Shanghai Education and Research Network, and Shanghai Telemedicine Center. Then, he joined SUN Microsystem Inc., Shanghai, China, as a Senior Member of the Technical Staff. He is currently an Assistant Professor at Nanyang Technological University, Singapore.

View full text