連豊力臺灣大學：電機工程學研究所李君威Li, Chun-WeiChun-WeiLi2007-11-262018-07-062007-11-262018-07-062004http://ntur.lib.ntu.edu.tw//handle/246246/52987傳輸控制協定之系統設計與效能分析 研究生:李君威 指導教授:連豊力 博士 國立臺灣大學 電機工程學系 摘要 在本篇論文中，我們首先介紹現今網路的架構，並且簡單的介紹每一層網路層的功用。接著我們專注於傳輸層的介紹。我們列出重要的TCP演算法，例如：TCP-Reno，以及AQM演算法例如: RED。我們使用非線性的動態離散時間方法去模型Reno/RED系統，這個系統視為非隨機式的回饋系統。首先我們對誤差動態向量矩陣模型線性化，以此來了解非線性系統的特性。同時討論回饋參數，例如:頻寬速度，使用者數目，來回時間等對系統造成的影響。 接著我們對所導出的動態模型做z轉換，可以得到系統的轉移函數。因此我們可以對轉移函數畫波德圖和尼可士圖來檢查系統的穩定度，並且使用特徵方程式找出系統極點的位置。根據極點的位置我們可以觀察出參數對系統所造成的影響。此外我們也使用控制系統中的參數：阻尼比以及自然頻率來觀察系統的表現。我們也改變初始條件，觀察在一般情況和不同初始條件下系統表現的差別。 最後我們檢查當參數改變對系統所造成的影響以及系統的變化當參數相對改變的情況。我們使用NS-2模擬器來驗證我們的分析及推論。 關鍵字：傳輸控制協定之雍塞控制，主動式流量管理，控制理論，非線性分析System design and performance analysis of Transmission Control Protocol with Active Queue Management Student: Chun-Wei Li Advisor: Feng-Li Lian Department of Electrical Engineering National Taiwan University ABSTRACT In this thesis, we first introduce current networks architecture and briefly explain the functions of each network layer. We mainly focus on the transport layer of the ISO/OSI seven layers. We list some important TCP algorithms such as TCP-Reno and AQM algorithms such as the RED algorithm. We look into the nonlinear dynamics in a discrete-time network model combined with Reno and RED. The system is regarded as the deterministic closed-loop system. We investigate its stability regions. First, we use linearized error dynamical vector-matrix model to present the nonlinear systems and discuss its feedback properties in terms of network parameters such as link capacity, number of senders, and round trip time. Second, we use the z-transform to transform the derived dynamical equations into the transfer function format. Thus, we can investigate the stability of the system by plotting Bode plot or Nichols chart of the transfer function and use the characteristic equation to find the poles of the system. Based on the location of the poles we can observe the parameter effects on the system performance. Besides, the pole location effects are also discussed by observing the damping ratio and natural frequency of the system. We also change the initial conditions to observe the difference in the normal case. Finally we investigate the equilibrium condition variations when the parameters are changed and the system performance on RED and network parameters are changed. We present NS-2 simulations to support our analysis. Key words: TCP congestion control, AQM, control theory, nonlinear analysisContents CHAPTER 1 1 INTRODUCTION 1 1.1 MOTIVATION 1 1.2 NETWORK LAYERED MODEL BACKGROUND 2 1.3 LITERATURE SURVEY OF RELATED RESEARCHES 5 1.4 CONTRIBUTION OF THE THESIS 14 1.5 ORGANIZATION OF THE THESIS 15 CHAPTER 2 16 TCP AND AQM ALGORITHMS & CONTROL ANALYSIS TOOLS 16 2.1 TCP-TAHOE ALGORITHM 16 2.2 TCP-RENO ALGORITHM 19 2.3 AQM-RED ALGORITHM 22 2.4 OTHER TCP AND AQM ALGORITHMS 24 2.5 BACKGROUND OF CONTROL THEORY 38 CHAPTER 3 48 ANALYSIS AND DESIGN OF NETWORK CONGESTION CONTROL 48 3.1 TCP CONGESTION MODEL RELATED WORKS 48 3.2 DISCRETE NONLINEAR DYNAMICS MODELS 51 3.3 A CONTROL THEORETIC ANALYSIS 56 3.4 POLE LOCATION EFFECTS AND SYSTEM OSCILLATIONS 60 CHAPTER 4 67 PERFORMANCE ANALYSIS 67 4.1 PARAMETER EFFECTS ON EQUILIBRIUM CONDITIONS 67 4.2 THE NETWORK SIMULATOR-NS-2 70 4.3 PARAMETER EFFECTS ON SYSTEM PERFORMANCES 72 CHAPTER 5 80 CONCLUSIONS AND FUTURE WORK 80 REFERENCES 82en-US傳輸控制協定之雍塞控制主動式流量管理非線性分析控制理論TCP congestion controlAQMcontrol theorynonli[SDGs]SDG11thesis