Implementing parallel shortest path for parallel transportation applications

Shortest path algorithms are required by several transportation applications; furthermore, the shortest path computation in these applications can account for a large percentage of the total execution time. Since these algorithms are very computationally intense, parallel processing can provide the compute power and memory required to solve large problems quickly. Therefore, good parallel shortest algorithms are critical for efficient parallel implementations of transportation applications. The experimental work related to parallel shortest path algorithms has focused on the development of parallel algorithms; however, very little work has been done with analyzing and understanding the performance impact of various implementation issues. In this paper, we conduct a thorough experimental analysis of parallel shortest path algorithms for sparse networks, concentrating on three implementation issues: (1) choice of shortest path algorithm, (2) termination detection and (3) network decomposition. The paper focuses on the choice of shortest path algorithm and network decomposition since the work on termination detection was published previously. We determine that all three issues affect the communication and convergence of the shortest path algorithm. Furthermore, we find that communicating the most information at a time results in the best convergence; this is contrary to most scientific applications where it is optimal to minimize communication.