| Numerous routing schemes have been reported to improve network performances over the years. Multi-path routing belongs to one of them and Multi Protocol Label Switching (MPLS) is an excellent platform for such routing. However, multi-path IP routing applied to the Internet has received little research attention because of the complexity of the problem and some intrinsic issues such as multi-objectives on path selection and traffic assignment on selected paths. In this thesis, an algorithm called Shortest Distance Path based Simulated Annealing (SDPSA) for finding optimal bandwidth reservation solutions for multi-path routing is developed to improve network performances. The algorithm, which employs the annealing method, is based on previous solutions to find the current sub-optimal solution for multi-path routing. Multiple objectives including balancing traffic load and minimizing network resource consumption are taken into consideration. The algorithm also integrates the interferences of multiple paths into IP routing optimization. Finally, the proposed algorithm is applied to a randomly generated network and the NSF network. The performance values are compared to a well-known multi-path routing algorithm-HSTwp. The simulation and comparison results show that the proposed SDPSA algorithm is efficient for the optimization of multi-path IP routing. |
