Evolutionary algorithms in MPLS network designing

MPLS technology become more and more popular especially in core networks giving great flexibility and compatibility with existing Internet protocols. There is a need to optimal design such networks and optimal bandwidth allocation. Linear programming is not time efficient and does not solve nonlinear problems. Heuristic algorithms are believed to deal with these disadvantages and the most promising of them are evolutionary algorithms. The problem is to define chromosome coding to reflect MPLS-specific path design and algorithms for finding paths, initialization of chromosomes population and genetic operators to be compatible with MPLS. The following paper is about tool that has been created to use evolutionary algorithm to design MPLS-based networks. The tool can generate paths and allocate bandwidth, on a basis of multiclass bandwidth demand matrix in a given network structure. It can also choose appropriate paths from a given pool and allocate bandwidth on them to achieve optimal network usage according to fitness function provided.