A simulation tool for tuning IP network parameters based on fluid-flow models and parallel genetic algorithms

摘要

In the current Internet scenario increasing interest is being shown in quality of service (QoS) issues, so simulation and parameter tuning are becoming very hot topics. In this paper we present a user-friendly, easily expandable and time-saving tool, which is able to perform network simulation through fluid-flow models and parameter tuning with parallel genetic algorithms. All the network elements involved (TCP sources, links, routers and so on) are modeled with the fluid-flow approach as independent systems: their outputs only depend on their inputs and internal state variables. Network elements can be connected to each other by a powerful graphical interface. Subsequently the user can select network parameters to be optimized in order to achieve prearranged QoS guarantees. During this phase the tool adopts a parallel implementation of a genetic algorithm (GA) which reduces the execution time required for optimization. A case study is also presented.