Doctor of Philosophy (PhD)
The study of complex problems in science and engineering today typically involves large scale data, huge number of large-scale scientific breakthroughs critically depends on large multi-disciplinary and geographically-dispersed research teams, where the high speed network becomes the integral part. To serve the ongoing bandwidth requirement and scalability of these networks, there has been a continuous evolution of different TCPs for high speed networks. Testing these protocols on a real network would be expensive, time consuming and more over not easily available to the researchers worldwide. Network simulation is well accepted and widely used method for performance evaluation, it is well known that packet-based simulators like NS2 and Opnet are not adequate in high speed also in large scale networks because of its inherent bottlenecks in terms of message overhead and execution time. In that case model based approach with the help of a set of coupled differential equations is preferred for simulations. This dissertation is focused on the key challenges on research and development of TCPs on high-speed network. To address these issues/challenges this thesis has three objectives: design an analytical simulation methodology; model behaviors of high speed networks and other components including TCP flows and queue using the analytical simulation method; analyze them and explore impacts and interrelationship among them. To decrease the simulation time and speed up the process of testing and development of high speed TCP, we present a scalable simulation methodology for high speed network. We present the fluid model equations for various high-speed TCP variants. With the help of these fluid model equations, the behavior of high-speed TCP variants under various scenarios and its effect on queue size variations are presented. High speed network is not feasible unless we understand effect of bottleneck buffer size on performance of these high-speed TCP variants. A fluid model is introduced to accommodate the new observations of synchronization and de-synchronization phenomena of packet losses at bottleneck link and a microscopic analysis is presented on different buffer sizes at drop-tail queuing scheme. The proposed model based methods promotes principal understanding of the future heterogeneous networks and accelerates protocol developments.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Kumar, Suman, "Model based analysis of some high speed network issues" (2010). LSU Doctoral Dissertations. 3382.