A machine learning based algorithm for routing bandwidth-guaranteed paths in MPLS TE: Improvements and performance assessment

The RRATE algorithm opened a new class of solutions for the computation of bandwidth-guaranteed paths in MPLS networks giving significant increase in both path selection speed and efficiency. Nevertheless, analyzing RRATE behavior while dealing with complex and big (in terms of nodes and link number) networks, has been shown that are possible more enhancements if changes are made to the core function that evaluates the “goodness” of a path over other paths. That function takes into account the critical links and the residual bandwidths available on the network. More specifically, changes have been made to the way those values are computed taking into account no more the scalar value of residual bandwidths and critical links but a ranking of the first values and an appropriate weighting of the second parameters. The obtained results reveal steady performance gains over the original RRATE algorithm and, in larger extent, over legacy algorithms like MIRA.