Performance Optimization and Dynamics Control for Large-scale Data Transfer in Wide-area Networks

Transport control plays an important role in the performance of large-scale scientific and media streaming applications involving transfer of large data sets, media streaming, online computational steering, interactive visualization, and remote instrument control. In general, these applications have two distinctive classes of transport requirements: large-scale scientific applications require high bandwidths to move bulk data across wide-area networks, while media streaming applications require stable bandwidths to ensure smooth media playback. Unfortunately, the widely deployed Transmission Control Protocol is inadequate for such tasks due to its performance limitations. The purpose of this dissertation is to conduct rigorous analytical study of the design and performance of transport solutions, and develop an integrated transport solution in a systematical way to overcome the limitations of current transport methods. One of the primary challenges is to explore and compose a set of feasible route options with multiple constraints. Another challenge essentially arises from the randomness inherent in wide-area networks, particularly the Internet. This randomness must be explicitly accounted for to achieve both goodput maximization and stabilization over the constructed routes by suitably adjusting the source rate in response to both network and host dynamics.The superior and robust performance of the proposed transport solution is extensively evaluated in a simulated environment and further verified through real-life implementations and deployments over both Internet and dedicated connections under disparate network conditions in comparison with existing transport methods