Queueing networks with blocking have proved useful in modeling stochastic service systems, such as computer and communication networks, manufacturing, traffic control, transportation, facility planning and many others. In this dissertation, we begin with studying the properties of a GI/G/1/N queue. We not only examine the characteristics of the queue length, busy periods, blocking probabilities, and transition probabilities between states, but also analyze properties of the general distributions. Based on the mathematical structures of the parameter space and solution space, a stochastic comparison method is utilized to classify the group of general distributions, and a diffusion process is utilized to approximate the blocking probability for a GI/G/1/N queue. The remainder of this dissertation is concerned with a generalized design methodology for open finite queueing networks. A hierarchical methodology is presented, in which large scale systems are divided into different subsystems, and dynamic programming is used to determine the decision variables, such as buffer size, for the systems with respect to the following two conflicting objectives: (1) Maximize average throughput; (2) Minimize processing cost. All theoretical results are tested by means of digital simulation and the different experiments are presented as an illustration of the proposed design methodology. Finally, open questions and directions for future research concludes the dissertation.
Identifer | oai:union.ndltd.org:UMASS/oai:scholarworks.umass.edu:dissertations-8960 |
Date | 01 January 1994 |
Creators | Wang, Xiaoyu |
Publisher | ScholarWorks@UMass Amherst |
Source Sets | University of Massachusetts, Amherst |
Language | English |
Detected Language | English |
Type | text |
Source | Doctoral Dissertations Available from Proquest |
Page generated in 0.002 seconds