臺灣博碩士論文加值系統

傳輸控制協定 (TCP) 是現今網際網路上最廣泛使用的傳輸層協定，它提供了網際網路端對端(end-to-end)間一個可以自我調適速度的可靠性傳輸，避免網路資源因過度被使用而造成的崩潰。然而傳統TCP Reno在壅塞避免階段，其壅塞視窗大小線性的增加過於緩慢，以致於當發生封包遺失的時候，無法在短時間內恢復高效能的傳輸率，所以在高速網路之下，效能不彰；隨著網際網路科技的迅速發展，網路的傳輸速度已經大幅成長，傳統的TCP反而成為高速網路上的瓶頸點。在本篇論文，為了解決這樣的問題，我們提出一個新的混合式TCP：Concave Convex TCP。CCTCP是藉由修改Reno的壅塞控制方法，增加了delay-based的特性來輔助增加頻寬使用率。它會根據網路的歷史資訊和網路現況來動態切換凹曲線成長方式和凸曲線成長方式，以調整它的壅塞視窗大小。除了利用ns-2網路模擬器模擬效能，並且我們實作CCTCP於現實的Linux系統平台上實測其效能，相較於其它混合式TCP壅塞控制協定，結果顯示在不同網路環境底下CCTCP皆能有較好的表現，特別在高速網路上，並且對於瓶頸佇列的緩衝區大小依賴程度相對上較低，而對於傳統TCP共存之下也有良好的公平性並增加整體效能使用率。

Transmission Control Protocol (TCP) is a widely used end-to-end transport protocol in the Internet. It is given the task to protect the Internet from collapse as well as to make a good use of network resources. However, the congestion control of standard TCP (Reno) functions poorly in high speed networks because of its slow response with large congestion windows. Therefore, standard TCP may become the performance bottleneck as the bandwidth of Internet continues to grow. In this paper, we propose a new variant of hybrid TCP, Concave Convex TCP (CCTCP), to overcome this issue. CCTCP revises Reno’s congestion avoidance phase by appending delay-based features. It dynamically switches its state between concave and convex states depending on historical records and network conditions to adjust its window size. Through ns-2 based simulations and experiments on the Linux platform we find that, as compared to other hybrid TCPs, CCTCP can achieve a high performance with minor dependence on bottleneck buffer sizes under a variety of network bandwidths, especially in high speed networks. Moreover, it coexists with standard TCP by higher degrees of TCP-friendliness.

Contents

Abstract in Chinese i
Abstract in English ii
Acknowledgements iii
Contents iv
List of Figures vi
List of Tables vii
1 Introduction 1


2 Related Works 3

2.1 Standard TCP (Reno) 4

2.2 Compound TCP (CTCP) 6
2.3 TCP-Illinois 9
2.4 TCP-Fusion 10

3 Concave Convex TCP (CCTCP) 11

3.1 The formula of dwnd 13

3.2 The calculations of α 13


4 Performance Evaluation 16

4.1 Basic behavior 17

4.2 Goodput performance 20

4.3 Fairness 22

4.4 Friendliness with standard TCP 23
4.5 Experiments on the Linux platform 25


5 Conclusions 28
Bibliography 29

List of Figures




2.1 The congestion window evolution in the CA phase of Standard TCP 5
2.2 The congestion window evolution in the CA phase of CTCP 6
2.3 Network topology for analysis 7
2.4 Congestion window behavior in CTCP with Buffer Size 100 packets 7
2.5 Congestion window behavior in CTCP with Buffer Size 25 packets 8

2.6 Additive increase α and multiplicative decrease β coefficients as a function
of queuing delay da 9

3.1
The concave and convex curve of CCTCP 12
3.2 The α calculation in the convex state 14

3.3 The switching between concave and convex states in CCTCP 15

4.1
Network topology for the simulations 17
4.2 Basic behavior with buffer size 0.1 BDP 18
4.3 Basic behavior with buffer size 0.2 BDP 19
4.4 RTT 50ms, TCP variants compete with Reno in different bottleneck band- widths 24
4.5 RTT 100ms, TCP variants compete with Reno in different bottleneck bandwidths 25
4.6 Network configuration used in the Linux based experiments 26


List of Tables





2.1 Goodput (Mbps) of Standard TCP and CTCP with different buffer size 8


4.1 The setting of simulation time 17

4.2 The goodput performance of basic behavior of Reno, CTCP, TCP-Illinois

and CCTCP with buffer size 0.1 BDP 18

4.3 The goodput of basic behavior of Reno, CTCP, TCP-Illinois and CCTCP

with buffer size 0.2 BDP 20

4.4 RTT 50 ms, performance of TCP variants with different flow numbers and bottleneck bandwidths 21
4.5 RTT 100 ms, performance of TCP variants with different flow numbers

and bottleneck bandwidths 21

4.6 RTT 50 ms, the Jain’s fairness index of TCP variants with 2 and 10 TCP

flows under different bottleneck bandwidths 22

4.7 RTT 100 ms, the Jain’s fairness index of TCP variants with 2 and 10 TCP

flows under different bottleneck bandwidths 23

4.8 Goodput of TCP variants with link delay 25 ms and different queue lengths 27

4.9 Goodput of TCP variants with link delay 50 ms and different queue lengths 27

Bibliography

Bibliography



[1] M. Allman, V. Paxson and W. Stevens, “TCP Congestion Control,” RFC 2581, April

1999.


[2] S. Floyd, “HighSpeed TCP for Large Congestion Windows,” RFC 3649, December

2003.


[3] T. Kelly, “Scalable TCP: improving performance in highspeed wide area networks,” ACM SIGCOMM Computer Communication Review, vol. 33, pp. 83-91, April 2003.

[4] D. Leith and R. Shorten, “H-TCP: TCP for high-speed and long-distance networks,”

in Proceedings of PFLDnet, 2004.


[5] L. Xu, K. Harfoush, and I. Rhee, “Binary increase congestion control for fast, long distance networks,” in Proceedongs of IEEE INFOCOM, vol. 4, pp. 2514-2524, March 2004.

[6] I. Rhee and L. Xu, “Cubic: A New TCP-friendly high-speed TCP variant,” ACM SIGOPS Operating Systems Review, vol. 42, pp. 64-74, July 2008.

[7] D. X. Wei, C. Jin, S. H. Low, and S. Hegde, “FAST TCP: motivation, architec- ture,algorithms, performance,” IEEE/ACM Transactions on Networking, vol. 14, pp. 1246-1259, December 2006.

[8] K. Tan, J. Song, Q. Zhang, and M. Sridharan, “A Compound TCP Approach for

High-Speed and Long Distance Networks,” in Proceedongs of IEEE INFOCOM, pp.

1-12, April 2006.


[9] S. Liu, T. Basar, and R. Srikant, “TCP-Illinois: a loss and delay-based congestion control algorithm for high-speed networks,” Performance Evaluation, vol. 65, pp.
417-440, June 2008.

[10] K. Kaneko, T. Fujikawa, Z. Su, and J. Katto, “TCP-Fusion: a hybrid congestion control algorithm for high-speed networks,” in Proceedings of PFLDnet, February
2007.


[11] W. Xiuchao, C. MunChoon, A. L. Ananda, and C. Ganjihal, “Sync-TCP: A new approach to high speed congestion control,” ICNP, pp. 181-192, October 2009.

[12] L. Brakmo and L. Peterson, “TCP Vegas: end to end congestion avoidance on a global Internet,” IEEE Journal on Selected Areas in Communication, vol. 13, pp.1465-1480, October 1995.

[13] A. Afanasyev, N. Tilley, P. Reiher, L. Kleinrock, “Host-to-Host Congestion Control for TCP,” CST, vol. 12, pp. 304-342, May 2010.

[14] The Network Simulator - ns-2. Available at http://www.isi.edu/nsnam/ns/


[15] R. Jain, D. Chiu, and W. Hawe, “A quantitative measure of fairness and discrim- ination for resource allocation in shared systems,” DEC Research Report TR-301, September 1984.

[16] L. Rizzo, Dummynet, Available at http://info.iet.unipi.it/~luigi/dummynet/

[17] The tcp/udp bandwidth measurement tool - Iperf. Available at http://
sourceforge.net/projects/iperf/