Enhancing path selection in multihomed nodes

Abstract

Path selection in multihomed nodes can be enhanced by optimization techniques that consider multiple criteria. With NP-Hard problems, MADM techniques have the flexibility of including any number of benefits or costs criteria and are open regarding the functions that can be employed to normalize data or to determine distances. TOPSIS uses the Euclidean distance (straight line) while DiA employs the Manhattan distance (grid-based) to determine the distance of each path to ideal values. MADM techniques have been employed in distinct areas, as well. Such openness and flexibility may lead to sub-optimal path selection, as their optimality is associated with functions that determine distance as a straight line or as grid path, and not inside an ideal range determined by the type of criteria. In this paper we propose the MeTH distance which considers the type of criteria, whether benefits or costs. In addition, we establish a MADM evaluation methodology based on statistical analysis that enables an objective comparison between MADM mechanisms and respective functions for path selection. With the proposed MADM evaluation methodology, we demonstrate that our MeTH distance is more efficient for the path selection problem than Euclidean and Manhattan distances.