Optimizing shortest path queries with parallelized arc flags

Abstract

Computing shortest path between nodes in a given directed graph is a very practical problem. Among the various shortest path algorithms, Dijkstra's shortest path algorithm is said to have better performance with regard to runtime. The speedup techniques when applied to shortest path algorithm were naturally improved the speedup of the algorithm there by the system performance. The combined speedup techniques are utilized along with Dijkstra's algorithm to further improve the performance in terms of runtime and number of vertices visited in the graph. The speedup techniques namely arc flags and bidirectional search were combined (COAB) and the speedup was measured with respect to runtime and number of nodes visited. The pre-processing phase of arc flags was parallelized to improve the performance of the system. The results were tested in random, planar and real world graphs.