An investigation and simulation of novel dynamic routing methods

Routing in networks is a multi-objective and multi-constraint optimization problem due to the nature of current networks being highly dynamic environments. Currently implemented solutions are single metric solutions where instead an optimal multi-metric solution is needed to solve this problem. This research work investigates novel multi-metric solutions to this optimization problem. Recently, it is found that the employment of a natural optimization process called the ant colony optimization process to the routing problem, resulted in a multi-metric dynamic solution. Latest research work reported two slightly different implementations of this employment. Network agents are used to sensor the status of the network and feedback the network nodes with the necessary information. This is used to update its routing tables based on the network status. These two implementations differ in the method (philosophy) used to update the information in the routing tables held by the network nodes. This research work suggests a new method to update the routing tables held by the nodes in the network. This done by merging modified versions of the previous two methods in order to overcome the disadvantages of each. A discrete event simulation system is built to test the routing method suggested by this research work together with the previous two routing methods for companson purposes. This simulation system represents a prototype for the development of a general network simulation tool. It is capable of collecting various types of simulation statistical data and generating tracing files for detailed studies of the network and for testing purposes. An expandable structured C-pointer based implementation is used to code the system. The system is tested on various networks and the results of the simulation show improvements on the performance of the network by reducing the overall delay in the network and increasing throughput. Moreover, the use of the suggested routing method results in balancing the load in the network.

Text FSKTM_2001_17 IR.pdf Download (1MB)

Actions (login required)

View Item