| One of the main objectives when designing a Multi-Service Internet is to define a profitable and efficient network configuration. Network profit is a very important optimization objective for today's Internet Service Providers (ISP). However, successful optimization for a large system such as a nation wide Internet backbone supporting both voice and data services requires a fast-responding and high-efficient method. In this thesis, we explore the methodology for such optimization design and present a two-phase method. In the first phase, a mathematical method based on a Linear Programming (LP) solution technique is utilized as primitive method with simplified objective functions and constraints. A Hopfield Neural Network (HNN) model is applied in the second phase to get an extended optimization for the network performance aiming to effectively distribute the network traffic across the network. The method has been developed and applied to the design of a national Internet backbone network. Both models are formally defined, optimally formulated, and solved. |
