摘要
随着计算机网络的快速发展,人们对网络资源的要求越来越高,尤其是近年来如语音、图像、视频等多媒体流在网络上的大量涌现,网络拥塞问题变得越来越严重。总的来说,网络产生拥塞的根本原因在于端系统提供给网络的负载大于网络资源容量和处理能力,表现为数据报延时、丢弃概率增加、上层应用性能下降等。为了确保Internet稳定性(stability)和鲁棒性(robustness),传统的TCP拥塞控制机制仍然使用最为广泛,并且占据着主导地位。但随着综合网络(WAN+LAN+3G蜂窝)特别是无线网络和高性能网络的崛起,要想在高误码率、大延迟、终端频繁移动、链路抖动、大容量数据交换的线路上仍然正常地运行现有的TCP控制协议,就变得相当的困难。因此要想在无线网络和高速网络中保持较高的通信效率、较高的吞吐量和可扩展性,改进现有的拥塞机制已迫在眉睫。
目前,对于无线网络和高性能网络的研究虽然尚处于初始阶段,但众多的科研工作者都已认识到问题的迫切性,进一步提高下一代网络拥塞控制机制的性能已经成为研究的热点。无线网络中,在大量研究的基础上发现TCP Westwood是一种较理想的算法,它的主要思想是通过在发送端持续不断的检测ACK的到达速率来进行带宽估计,当拥塞发生时用带宽估计值来调整拥塞窗口和慢启动阈值,采用AIAD(Additive Increase and Adaptive Decrease)拥塞控制机制。它不仅提高了无线网络的吞吐量,而且具有良好的公平性和与现行网络的互操作性。存在的问题是不能很好的区分传输过程中的拥塞丢包和无线丢包,导致拥塞机制频繁调用。高性能网络中综合表现比较优秀的算法是:H-TCP,但它有RTT不公平性和低带宽不友好性等问题。对于中间结点算法,新提出的BLUE给我们带来了很大惊喜,然而也有队列长度抖动大和参数设置敏感等问题。
本文首先介绍当前宽带网络拥塞的研究状况,论述RTT的相关估计方案和测量方法,给出相应的评价准则,根据相关背景提出研究方向。接着重点介绍自己的研究成果,主要如下:
第一,针对TCPW在运行过程中不能很好的区分拥塞丢包和无线丢包,导致拥塞机制调用频繁,降低利用率等问题,提出了TCPW BR算法。该算法改善RTT评估体系,利用超时重传定时器的方法来预测平滑的往返延迟值,将测得的RTT值带入加权平均公式F = RTTmax ? RTTmin;R = (RTT-RTTmin)/F。根据R值的大小来划分拥塞等级,区分无线丢包和拥塞丢包。同时利用反馈因子根据划分的等级结合原有的机制共同应对拥塞问题。仿真结果表明TCPW BR算法提高了带宽利用率,保持了良好的公平性和友好性。
第二,针对高性能网络中端算法H-TCP和中间结点算法BLUE中存在的问题,提出了新的TCP/AQM模型的高带宽拥塞算法,称之为H-TCP*/BLUE*算法。该算法将源算法和中间结点算法相结合,通过添加公平性因子n=BWE/BWEmax*RTT,在中间结点设置阈值和划分调控步长,利用显示反馈等思想来改善拥塞机制。仿真结果表明,新的算法无论是在带宽利用率还是在公平性、友好性方面都有显著提高。
最后,给出论文的总结,就怎样改善RTT的评估方案和应用数学模型做一下初步探讨。给出一些新思路新想法,以便在今后的工作中继续深入研究。
With the rapid development of computer networks, people have become increasingly demanding of network resources, especially in recent years, such as voice, video, video streaming and other multimedia on a large number of the network, network congestion has become a serious problem. In general, network congestion arising from the fundamental reason for this is provided to the network end system load is greater than the capacity of network resources and capacity to deal with, the performance data reported for the delay, increase the probability of discarding the upper drop in application performance. In order to ensure stability and robustness, the traditional TCP congestion control mechanism is still the most widely used and occupied a dominant position. However, with the rise of the integrated network (WAN + LAN +3 G cellular)especally wireless networks and high-performance networking,.it is rather difficult to run the existing TCP control protocol if on the high bit error rate, large delays, frequent mobile terminal, link jitter, high-capacity data exchange line. Therefore in order to maintain a high efficiency, high throughput and scalability on high-speed wireless networks and network communications , improving the existing mechanisms is imminent congestion.
At present, the wireless network and high-performance network research while still at the initial stage, but many scientists have come to understand the urgency of the issue . It has become a study of the important performance of further enhancing the next generation of network congestion control mechanism .In the wireless networks, we find that TCP Westwood is a kind of found a better algorithm in the basic many research, it’s main idea that at the sending end through the continuous detection of the arrival rate of ACK for bandwidth estimation, congestion occurs when the bandwidth is estimated using value to adjust the congestion window and slow start threshold, the AIAD (Additive Increase Adaptive Decrease) congestion control mechanisms. It not only improve the throughput of the wireless network, but also has good fairness and with the existing network interoperability. The problem should not exist are a very good transfer process to distinguish between congestion loss and wireless loss, resulting in frequent call congestion mechanism. General performance of high-performance network comparison algorithm is excellent: H-TCP, but it has problem that RTT unfairness and low bandwidth unfriendly. Algorithm for node between the new proposed BLUE give us a big surprise, but there is also the queue length jitter and the sensitive issues of parameter settings. This paper first introduces the current research broadband network congestion situation, discussed the relevance of the estimated RTT and measurement method, give the appropriate evaluation criteria, in accordance with the relevant background research. Then focuses on the results of their research, are as follows:
First, in the course of TCPW should not run at a very good distinction between congestion packet loss and wireless packet loss, resulting in a mechanism called the frequent congestion, reducing the utilization rate and so on. TCPW BR proposed algorithm to improve the RTT evaluation system, the use of retransmission timer time-out method to smooth the prediction values from the delay will be measured by the weighted average of RTT values into the formula F = RTTmax - RTTmin; R = (RTT -RTTmin) / F. According to the size of R, the value of congestion level is divided to distinguish wireless loss and congestion loss. At the same time, use the feedback factor in accordance with the level of integration into the mechanism previously common problem to deal with congestion. The simulation results show that the TCPW BR algorithm improve bandwidth utilization, and maintain a good fairness and friendliness.
Second, in the course of H-TCP algorithm of high-performance network and BLUE algorithm of middle node between the problems that exist,we find a new TCP / AQM model of high-bandwidth congestion algorithm, known as H-TCP * / BLUE * algorithm. The algorithm is a combination algorithm of the source node algorithm and middle algorithm , the fairness factor by adding n = BWE / BWEmax*RTT, set up in the middle node control threshold and step into the use of display, meanwhile the feedback mechanisms also improve congestion control. The simulation results show that the new algorithm are in the bandwidth utilization regardless of and in equity, friendliness have improved significantly.
Finally, give a summary of thesis, on how to improve the RTT application of the assessment program and a preliminary mathematical model to explore what to do. Give some new idea that in order to work continue to develop in the future.
引文
[1]罗万明,林闯,阎宝平. TCP/IP拥塞控制研究.计算机学报, 2001, Vol.24, No.1, 1~18
[2] Jain, R. Congestion Control in Computer Networks: Issues and Trends. IEEE Network Magazine, 1990, 4(3):24~30
[3] CASETH C,GERLA M,MASCOI O S,et a1.TCP Westwood:End- to-End Bandwidth Estimation for Enhanced Transport over Wireless Links[J].Journal of Wireless Networks,2002.467—479.
[4] S.Floyd. RFC 3649. HighSpeed TCP for Large Congestion Windows [S]. IETF, December 2003.
[5] T Kelly. Scalable TCP: improving performance in high-speed wide area networks [J].ACM Computer Communications Review, 2003, 33(2):83-91..
[6] Injong Rhee, and Lisong Xu, CUBIC: A New TCP-Friendly High-Speed TCP Variant [EB/OL], http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/cubic-paper.pdf,2004.
[7] Leith D, Shorten R. H-TCP: TCP for high-speed and long-distance networks. [EB/OL]. (2004). http://www.hamilton.ie/net/research.htm.
[8] Floyd, S. A Report on Some Recent Developments in TCP Congestion Control. IEEE Communication Magazine, 2001, 39(4):84~90
[9] Feng W, Kaudlur D, Saha D, Shin K. Blue: A New Class of Active Queue Management Algorithms. Technical Report CSE-TR-387-99, University of Michigan, 1999.
[10] V. Misra, W. Gong, D. Towsley, A fluid-based analysis of a network of AQM router supporting TCP flows with an application to Red [A], In: ACM SIGCOMM 2000, Stockholm, Sweden, September 2000
[11] Sanjeewa A, Steven H.Low. REM: Active Queue Management. IEEE Network, 2001,15(3):48~53.
[12] Allman, M., Floyd, S., Partridge, C. Increasing TCPs Initial Window. RFC 2414, 1998.
[13] Visweswaraiah, V., Heidemann, J. Improving Restart of idle TCP Connections. Technical Report, USC TR 97-661, University of Southern California, 1997. http://www.isi.edu/~johnh/PAPERS/.
[14]章淼,吴建平,林闯.互联网端到端拥塞控制研究综述[J].软件学报.2002,3:354-362.
[15] Andrew S. Tanenbaum(原著)潘爱民(译).计算机网络[M].北京:清华大学出版社, 2004.
[16] Alexander G. Parlos. Identification of the Internet End-to-end Delay Dynamics Using Multi-Step Neuro-Predictors. Neural Networks, 2002. Proceedings of the 2002 International Joint Conference on, Spain, 2002, 3:2460-2465.
[17] Anindya Basu, S. Jamaloddin Golestani. Estimation of Receiver Round Trip Times in Multicast Communications. RMRG Meeting, Virginia, AR, Dec. 1998:1一10.
[18] Yolanda Tsang, Mehmet Yildiz, Paul Barford et al. Network Radar Tomography from Round Trip Time Measurements. ACM IMC, Taormina, Sicily, Italy, October, 2004:175-180.
[19] Zhen Zhu, Albert M. K. Cheng. Worst Case Round Trip Time Prediction and Statistical Analysis Using Extreme Values Theory. Proc. WIP Session of IEEE-CS Real-Time Systems
[20] Claudio Casetti, Mario Gerla, et al. TCP Westwood: End to End Bandwidth Estimation for EnhancedTransport over Wireless Links[C]. Rome, Italy: ACM Mobilecom, 2001:287-297.
[21]张莉,张兴会.基于RTT的智能预测方法研究[J].天津工程师范学院学报.2007.3 ,1:34-36.
[22]张轶博,雷振明.一种被动式RTT测量算法[J].北京邮电大学学报[J].2004.10 ,4:85-89.
[23] Wang JunFeng, Zhou MingTian, Li YuXia. Survey on the End-to-End Internet Delay Measurements. Z. Mammeri and P. Lorenz (Eds.):HSNMC 2004, LNCS 3079, 2004:155-166.
[24] Srinivas Shakkottai, R. Srikant, NeviI.Brownlee et. al. The RTT Distribution of TCP Flows in the Internet and its Impact on TCP-based Flow Control. CAIDA Tech Report, number tr-2004-02, February 2004:1-15.
[25]卢利琼.拥塞控制端算法研究[D].广西师范大学.2006
[26] FP Kelly, A.Maulloo and D. Tan. Rate control in communication networks:shadow prices, proportional fairness and stability. Journal of the Operational Research Society, 1998,49:237-252.
[27] F.Kelly. Charging and rate control for elastic traffic. Eur.Trans.Telecommun, 1997, 8: 33-37.
[28] Jacobson V. Congestion Avoidance and Control [J]. IEEE/ACM Trans Networking, 1988, 6(3):314-329.
[29] Handley H .Floyd S ,PadhyeJ ,et al .RFC 3448 TCP Frien-dly Rate Control( TFRC) Protocol Specification[S].[s.1.]: [s.n.],2003.
[30]刘壮礼,陈元琰,罗晓曙,闫友彪.混合无线链路高速网络中一种基于TCPW的拥塞控制改进算法[J].计算机应用,2005,47-49.
[31]文远保,石正贵.无线网络的拥塞控制机制研究[J].计算机工程与科学,2004,27-29
[32]徐昌彪,隆克平,杨士中.无线网络下一种改进的TCP拥塞控制机制TCP_RD.通信学[J],2003,86-90.
[33] Arjan Duresi·Vamsi Paruchuri. Adaptive backbone protocol for heterogeneous wireless networks[J]. Telecommun Syst (2008) 38: 83–97
[34] Paxson V ,Allman M. Computing TCP’s Ret ransmission Timer[M] . RFC2988 ,2000
[35]杨新宇,曾明,江晓,赵瑞,吴航.一种新的自适应网络拥塞控制算法[J].计算机工程, 2004, 17-18.
[36]张瑞,洪佩琳,李津生,郭立鹏无线网络中一种改进的TCP拥塞控制机制[J].电路与系统学报,2006.7-13
[37]徐雷鸣,庞博,赵耀.NS与网络模拟[M].北京:人民邮电出版社, 2003.
[38]汪小帆,孙金生,王执铨,控制理论在Internet拥塞控制中的应用,控制与决策,2002,V.17(2):129-134
[39] Dina Katabi,Mark Handley, Charlie Rohrs. Internet Congestion Control for Future High Bandwidth-Delay Product Environments [DB/OL ]. http://www.ana.lcs.mit.edu/dina/XCP/, 2002-5/2004-3-10
[40]陈元琰,卢利琼,罗晓曙.一种基于H- TCP的拥塞控制改进算法[J].计算机工程与应用,2007, 43( 18). 115-116
[41]苏聪,陈元琰,罗晓曙,吴东.基于模糊理论的主动队列管理算法—FBLUE[J].计算机工程与应用,2006.23. 117-120.
[42] Steven H. Low and David E. Lapsley. Optimization Flow Control—I: Basic Algorithm and Convergence. IEEE/ACM Transactions on Networking, 7(6):861–874, 1999.
[43] Steven H. Low and David E. Lapsley. Optimization Flow Control—II: Implementation[DB/OL]. http://netlab.caltech.edu,2001-09-15/2001-10-25
[44] Low S H. A Duality Model of TCP and Queue Management Algorithms [DB/OL ]. http://netlab. caltech. edu, 2003-8/2004-3-10
[45] Paganini F. On the Stability of Optimization-based Flow Control [DB/OL ]. http:// www.ee.ucla.edu/ ~paganini, 2001-9/2004-3-10
[46] Kelly F P. Mathematical Modelling of the Internet [DB/OL ]. http://www.statslab.cam.ac.uk/ ~frank/mmi.html, 2001-9-20/2004-3-10
[47] Gibbens R J, Kelly F P. Resource pricing and the evolution of congestion control[J]. Automatica, 1999, 35(12): 1969-1985.
[48] Cheng Jin David X. Wei Steven H. Low. FAST TCP: Motivation, Architecture, Algorithms, Performance [DB/OL]. In: the Proceedings of IEEE Infocom, Hong Kong, March 2004, http://netlab.caltech.edu/fast, 2004-3/2004-3-20.