The Pennsylvania State University The Graduate School College of Engineering NETWORK QUEUE MANAGEMENT AND CONGESTION CONTROL IN INTERNET AND WIRELESS NETWORKS

Rapid development of telecommunication technologies and the ever growing network users demands have made network congestion a prominent problem in today’s Internet. Congestion not only brings significant performance degradation to the network, but also hurts the Quality-of-Service (QoS) that the users receive and even raises up some security concerns such as Denial-of-Service (DoS) attacks. The overall thesis is to investigate the role of flow control mechanisms in today’s Internet congestion control and QoS provisioning through extensive design and analysis. This dissertation studies several related research issues. First, we compare the performance of various Active Queue Management (AQM) schemes in providing QoS guarantees from a network-centric standpoint such as link throughput and router queue length. We propose a new AQM scheme, called Stabilized Virtual Buffer (SVB), that provides better performance and stability under a wide range of traffic and network conditions. To have a better understanding of the behavior of the AQM schemes, we then investigate the interaction between an AQM scheme and the end-system congestion control mechanisms using a control-theoretic approach and develop a generalized framework for analyzing various AQM schemes. Based on this model, we propose a generic implementation of AQM schemes as proportional-integral-derivative (PID) controllers. We also propose an adaptive controlling mechanism into the PID model to improve the system stability under changing network conditions. Next, we study how the AQM schemes behave in providing QoS assurance from a network user’s perspective. We investigate various techniques for extending the existing AQM schemes to address the fairness issues in a heterogeneous network. The above mentioned congestion control, fairness and security issues are not only prominent for the Internet infrastructure when wired backbones and centralized control entities are available, but also are very essential for various applications that use wireless technologies for flexible and reliable data communication. Therefore, this dissertation extends the research methodologies and technologies that we developed in the study of Internet congestion control problem to the various types