Date of Award
Master of Science
Department of Electrical and Computer Engineering
Richard A. Raines, PhD
In wide area computer data communications, many networks have evolved by satisfying increased user demands in the most expedient manner. In some cases, new users’ demands are satisfied by installing a new link, rather than sharing the links that are already in place. This research investigates the differences in performance between using a dedicated link for each source destination pair (nonshared bandwidth) and using a single link to be used by all source destination pairs (shared bandwidth). Simulation models are developed for a wide area network using shared bandwidth, and a wide area network using nonshared bandwidth. The quality of service offered by each network is based on its responsiveness and productivity. Responsiveness will be measured in terms of average end to end delay of packet transmission, and productivity will be measured in terms of percent bandwidth utilization. The networks are modeled under a common set of operating assumptions and system environment. This allows for accurate comparison of packet delay and bandwidth utilization. Two variable input parameters are used in the simulation: intensity of input traffic load, and amount of link capacity. Provided that the intensity of the input traffic load remains below the network saturation level, it is shown that the shared system clearly outperforms the nonshared system. This result occurs for both a uniform and a nonuniform traffic load destination distribution.
DTIC Accession Number
Stevens, John P., "Performance Study of Shared Versus Nonshared Bandwidth on a Packet-Switched Network" (1996). Theses and Dissertations. 6223.