Parallel Algorithm for Single-Source Earliest-Arrival Problem in Temporal Graphs

Many real-world networks, including online social networks and communication networks, are commonly modeled as temporal graphs. Answering earliest-arrival queries in temporal graphs is one of the most fundamental studies with numerous applications, such as information diffusion and measuring temporal closeness centrality. As graph sizes are growing rapidly, speedup of query execution time becomes even more important.In this paper, we propose a novel edge-centric parallel algorithm for solving single-source earliest-arrival problem in temporal graphs based on a new data structure named Edge-Scan-Dependency Graph (ESD-Graph). We evaluate the proposed parallel algorithm by theoretical analysis as well as by empirical experiments on real-world temporal graphs and synthetic graphs. Empirical results show that the new parallel algorithm outperforms the existing serial algorithm by up to 8.2 and 9.5 times on multi-core processors for real-world data and synthetic data respectively.