Control Methods for Data Flow in Communication Networks
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the
Graduate School of The Ohio State University
By
Peng Yan, M.S.
*****
The Ohio State University
2003
Dissertation Committee: Approved by
Professor Hitay Ozbay¨ , Adviser Professor Kevin M. Passino Adviser ¨ ¨ Professor Umit Ozg¨uner Department of Electrical Engineering c Copyright by
Peng Yan
2003 ABSTRACT
In this dissertation, we investigate various control methods for data flow in communica-
tion networks. First, we develop a rate-based flow controller for self-similar network traffic, ½ which consists of a robust À control block and an adaptive LMMSE capacity predictor.
The controller guarantees robust stability against time-varying time delay uncertainties and improves the transient response by predicting the self-similar cross traffic. Window-based congestion control methods are also explored for TCP traffic on IP networks. We propose a variable structure approach in Active Queue Management (AQM) support Explicit Con- gestion Notification (ECN). By analyzing the robustness and performance of the control scheme for the nonlinear TCP/AQM model, we show that the proposed design has good performance and robustness with respect to the uncertainties of the round-trip time (RTT)
and the number of active TCP sessions, which are central to the notion of AQM. Alterna- ½
tively, we design robust À AQM controllers for the linearized TCP/AQM model, with ½
the presence of uncertain time delays. The À performance is analyzed and a switching
control scheme is introduced to improve the system performance. Motivated by the Lin- ½ ear Parameter Varying (LPV) nature of the linearized TCP/AQM model, a switching À
control method is further investigated for LPV systems where we provide some stability
conditions in terms of the dwell time and the average dwell time.
ii To my wife, Yang Sun
iii ACKNOWLEDGMENTS
I would like to thank my adviser, Professor Hitay Ozbay,¨ for his guidance, support, and encouragement throughout the course of my research at The Ohio State University. I also thank Professor Kevin Passino and Professor Umit¨ Ozg¨¨ uner for serving on my committee and for their review of this thesis.
I would like to thank Dr. Yuan Gao for his collaboration on the topic of the Variable
Structure AQM in Chapter 3 and for providing me with the ns-2 code in the packet-level simulations of Chapter 3. I also thank Dr. Pierre-Franc¸ois Quet for various technical dis- cussions.
A special thanks goes to the entire Control group, professors and students, for creating a wonderful academic environment that encourages learning and pursuit of research. I would like to express my gratitude to all my friends for their help and encouragement.
Last but not least, I wish to specially thank my family for their love, support and en- couragement throughout my studies.
Finally I would like to acknowledge that the financial support for this work came form
NSF grants Nos. ANI-9806660, ANI-0073725, SBC/Ameritech Faculty Research Grant, and Air Force Research Laboratory under agreement No. F33615-01-2-3154.
iv VITA
January 21, 1975 ...... Born - P. R. China
1997 ...... B.S. Southeast University, Nanjing, P. R. China 1999 ...... M.S. Southeast University, Nanjing, P. R. China 1999-present ...... Graduate Research Associate, The Ohio State University.
PUBLICATIONS
Research Publications
Peng Yan, Yuan Gao, and Hitay Ozbay¨ Variable Structure Control in Active Queue Man- agement for TCP with ECN. Proceedings of the 8th IEEE Symposium on Computers and Communications, Antalya, TURKEY, July 2003.
¨ ½ Peng Yan and Hitay Ozbay À Performance Analysis of Robust Controllers Designed for AQM. Proceedings of the American Control Conference, Denver, Colorado, USA, June 2003.
¨ ½ Peng Yan and Hitay Ozbay On the À -Based Controllers for Automatic Steering of Vehi- cles with Actuator Delays. Proceedings of the 6th ASME Biennial Conference on Engieer- ing Systems Design and Analysis, Istanbul, Turkey, July 2002.
Peng Yan and Hitay Ozbay¨ Flow Controller Design and Performance Analysis for Self- Similar Network Traffic. Proceedings of the 39th Annual Allerton Conference on Commu- nication, Control, and Computing, Monticello, Illinois, USA, October 2001.
v FIELDS OF STUDY
Major Field: Electrical Engineering
vi TABLE OF CONTENTS
Page
Abstract ...... ii
Dedication ...... iii
Acknowledgments ...... iv
Vita...... v
List of Figures ...... x
List of Tables ...... xiii
Chapters:
1. Introduction ...... 1
1.1 Congestion Control in Communication Networks ...... 2 1.1.1 Rate-Based Control Methods for ATM Traffic ...... 3 1.1.2 Window-Based Congestion Control Methods for TCP Traffic . . 4 1.1.3 Active Queue Management ...... 5 1.1.4 Stability of Distributed Congestion Control ...... 7 1.2 Problem Definition and Motivation ...... 8 1.3 Contributions of the Dissertation ...... 9 1.4 Structure of the Dissertation ...... 11
2. Flow Controller Design and Performance Analysis for Self-Similar Network Traffic...... 12
2.1 Overview ...... 12 2.2 Preliminary definitions ...... 14 2.3 Mathematical Model ...... 15
vii 2.4 Short-Term Prediction of Self-Similar Traffic ...... 19 2.5 Simulation Results of the LMMSE Predictor ...... 20 2.6 System Level Simulations ...... 25 2.7 Concluding Remarks ...... 29
3. A Variable Structure Control Approach to Active Queue Management for TCP with ECN ...... 30
3.1 Overview ...... 30 3.2 Variable Structure Control in AQM ...... 32 3.2.1 Nonlinear TCP dynamics ...... 32 3.2.2 VS based AQM with ECN ...... 34 3.3 Robustness and Performance Analysis ...... 36 3.4 Related Work ...... 42 3.5 Packet-Level Simulations ...... 47 3.5.1 Simulation Configuration ...... 47 3.5.2 The Scenario of Single Bottleneck Topology ...... 48 3.5.3 The Scenario of Multiple Bottleneck Topology ...... 55
3.6 Concluding Remarks ...... 58 ½
4. Robust Controller Design for AQM and À -Performance Analysis ...... 61
4.1 Introduction ...... 61
4.2 Mathematical Model of TCP/AQM ...... 62 ½
4.3 À Controller Design for AQM ...... 64
4.4 Multiplicative Uncertainty Bound ...... 66 ½
4.5 À -Performance Analysis ...... 69 4.6 Simulations ...... 72 4.6.1 The Case of a Single Controller ...... 73 4.6.2 The Case of Switching Control ...... 75
4.7 Conclusions ...... 76 ½
5. On Switching À Controllers for a Class of LPV Systems ...... 79
5.1 Introduction ...... 80 5.2 Problem Definition ...... 81 5.3 Main Results ...... 84 5.4 Numerical Example ...... 92 5.5 Concluding Remarks ...... 99
viii 6. Conclusions ...... 100
6.1 Summary of Results ...... 100 6.2 Future Work ...... 102
Bibliography ...... 104
ix LIST OF FIGURES
Figure Page
1.1 The flow control model ...... 4
1.2 Data flow in communication networks ...... 8
2.1 LMMSE based feedback control system ...... 16
2.2 One-step prediction for different values of ...... 21
2.3 Error for different prediction steps ...... 22
2.4 Prediction Error and relative error variance in one-step prediction ...... 23
2.5 Discretized model ...... 25