Parallelization of Shortest Path Class Algorithms: A Comparative Analysis

Abstract

The problem of finding the shortest path between a source and a destination node, commonly represented by graphs, has several computational algorithms as an attempt to find what is called the minimum path. Depending on the number of nodes in-between the source and destination, the process of finding the shortest path can demand a high computational cost (with polynomial complexity). A solution to reduce the computational cost is the use of the concept of parallelism, which divides the algorithm tasks between the processing cores. This article presents a comparative analysis of the main algorithms of the shortest path class: Dijkstra, Bellman-Ford, Floyd-Warshall and Johnson. The performance of each algorithm was evaluated considering different parallelization approaches and they were applied on general and open-pit mining databases present in the literature. The experimental results showed an improvement in performance of about 55% on the execution time depending on the chosen parallelization point.