Techniques for quality of service improvement in Internetworks.
Files
Date
2014
Authors
Zhou, Jianghe
Editors
Advisors
White, Langford Barton
Ng, Brian Wai-Him
Sorell, Matthew James
Ng, Brian Wai-Him
Sorell, Matthew James
Journal Title
Journal ISSN
Volume Title
Type:
Thesis
Citation
Statement of Responsibility
Conference Name
Abstract
Quality of service in the Internet has long been a concern of computer network researchers. TCP congestion control is one of critical issues in achieving a good quality of service in Internetworks, for TCP which is able to provide a reliable data transmission over Internet is a core of modern Internet protocols suite. TCP uses a “window size” variable to limit the maximum number of packets that can be outstanding. TCP congestion control research focuses on design of congestion control algorithms which dynamically adjust congestion window (Cwnd) size and improve quality of service in terms of higher network throughput, smaller packets time delay and better fairness in network resource allocation to network users. This thesis presents design of new TCP congestion control algorithms as applications of modern control theory and game theory. The thesis consists of three primary parts: TCP congestion control background research, TCP congestion controller based on control theory and game theoretic TCP congestion control algorithm. The thesis first conducts background research in TCP congestion control including review and comparison of various TCP congestion control algorithms proposed by researchers. The background research shows that TCP Vegas is the most appropriate algorithm which is chosen as a foundation of research work in this thesis, for it is relatively easy to build control theoretic model and game theoretic model based on TCP Vegas for new algorithms design due to unique functions and parameters in TCP Vegas. A linear feedback system model Vegas for TCP previously developed by White and Chiera provides a foundation upon which the new algorithms and methodologies of this thesis are developed. The first problem addressed is a H∞ congestion controller based on modification of TCP Vegas as a robust generalisation of the H₂ approach of White and Chiera. The thesis then proposes a centralised TCP congestion controller as an application of Linear Quadratic Gaussian (LQG) robust control method. A centralised control model is established based on TCP Vegas by using the linear approach of White and Cheira, in which Cwnd of users are control inputs and packets time delays are control measurements. A LQG controller which consists of a Kalman filter and LQG regulator is designed to provide an optimal control for the control model. The primary aim of this work is to investigate potential performance benefits which might be gained by users in a sub-network sharing information about their TCP congestion control states. Naturally following the investigation of co-operation from a control viewpoint, is an approach based on game theory. The thesis makes a significant contribution to the development of adaptive mechanisms in repeated games. The thesis develops a new weighted regret matching (WRM) controller for TCP congestion control using the game theoretic approach. A WRM approach leading to a set of correlated equilibria for noncooperative games is proposed, which uses regret functions to update the probability mass function on users’ actions set and leads the probability distribution of users’ global behavior converge to a set of correlated equilibria. A TCP congestion control game model based on TCP Vegas is established and WRM approach is applied to the game model in which network users choose their Cwnd according to probability distribution function suggested by WRM algorithm. The parameters setting of WRM controller is discussed and the performance of proposed controller in TCP congestion control is investigated based on a single bottle-neck queue network simulated in Matlab. The compatibility and fairness of WRM controller is investigated in NS-2, a widely used network simulation software for network research, to provide a realistic network environment for this research.
School/Discipline
School of Electrical and Electronic Engineering
Dissertation Note
Thesis (Ph.D.) -- University of Adelaide, School of Electrical and Electronic Engineering, 2014
Provenance
This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at: http://www.adelaide.edu.au/legals