Traffic analysis, performance evaluation and enhancement in modern data communication networks
MetadataShow full item record
Traffic characteristics, which include both short range and long range characteristics, are one of the most important factors that affect the performance of modern data communication networks. Although during the last few years, significant research results have been proposed on models that capture long range dependency in traffic and its impact on Internet performance, they are inadequate for predicting queuing processes, delays and loss rates under dedicated scheduling disciplines or within specific network systems such as optical networks and wireless networks. This work first carries out traffic characteristic studies in communication networks with specific system details. The network systems under study include general communication networks with specific scheduling discipline--Generalized Processor Sharing (GPS), high-speed optical networks with Optical Burst Switching (OBS) paradigm and data intensive wireless cellular networks. Together with traffic characteristic studies, congestion control (using, e.g., TCP, rate assignment or control) is another important factor affecting the network performance, which is studied in this work. In particular, we consider a bufferless network core (as in the case of optical networks) and study how to improve the performance of TCP by enhancing its congestion mechanism. In addition, we investigate how to take advantage of the knowledge of data traffic characteristics to improve rate control and admission control mechanisms in wireless cellular networks. We aim at a comprehensive approach to studying traffic characteristics and enhancing the performance of congestion control mechanisms in different network systems. We develop traffic models and algorithms taking into consideration of network specifics such as schedule disciplines, transmission mechanisms, buffer availability and wireless channel conditions involved. The obtained results and implementations are evaluated analytically and via examples and simulations.