SP Async:Single Source Shortest Path in Asynchronous Mode on MPI

TL;DR

This paper introduces a novel MPI-based asynchronous algorithm for the single source shortest path problem that combines intra-node Dijkstra with inter-node Bellman-Ford, optimizing large graph processing.

Contribution

It presents a new hybrid MPI algorithm with triangle-based edge pruning and two termination detection methods for efficient large-scale SSSP computation.

Findings

Improved performance on large graphs.

Effective termination detection techniques.

Hybrid intra- and inter-node approach.

Abstract

Finding single source shortest path is a very ubiquitous problem. But with the increasing size of large datasets in important application like social network data-mining, network topology determination-efficient parallelization of these techniques is needed to match the need of really large graphs. We present a new Inter node-bellman cum Intra node Dijkstra technique implemented in MPI to solve SSSP problem. We have used a triangle based edge pruning for idle processes, and two different techniques for termination detection. Within each node the algorithm works as Dijkstra and for outer communication it behaves as inter node bellman ford. First termination detection technique is based on the token ring and counter. Second is a heuristic based technique, in which the timeout is calculated from the number of inter-edges and number of partitions. In this project asynchronous mode of message passing is used.