Abstract
This research investigates the performance and design of integrated circuit/packet-switched computer-communication networks. The integrated network under consideration has a circuit-switched communications subnet whose trunk lines carry both voice and data simultaneously. Peripheral packet switches provide data subscribers access to the subnet, while voice subscribers terminate directly to the circuit switch nodes of the subnet. An existing simulation model is modified and subsequently used to analyze the performance of integrated networks. A simulation experiment is designed and conducted to identify key network parameters and to determine the relationships that exist between these parameters and network performance. Multiple regression analyses are conducted to obtain models that describe the mean behavior of certain network performance measures with respect to changes in network traffic load, link capacity, and network size. The exact solution to the integrated network topology design problem is seen to be intractable for even small networks. This dissertation addresses the topology design problem using an iterative, heuristic approach whereby many suboptimal solutions (local minima) are generated in lieu of one optimal solution. The iterative scheme integrates the simulator as a network performance generation device into a performance feedback loop which dynamically reconfigures the network topology. An integrated cut-saturation add heuristic and a reliability-preserving delete heuristic are developed to move the network topology in the direction of an optimal, feasible solution. The topology design methodology is implemented as a modularized FORTRAN program and is applied to several networks of varying design specifications. Results demonstrate that the methodology developed constitutes a viable tool that can be used to analyze, design, and modify the integrated networks of the future.
Kiemele, Mark Ja (1984). Adaptive topological configuration of an integrated circuit/packet-switched computer network. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -554161.