Parallel Hub Labeling Maintenance With High Efficiency in Dynamic Small-World Networks

Shortest path computation is a fundamental operation in many application domains and is especially challenging in frequently evolving small-world networks (i.e., graphs in which many nodes can be reached from every other node by a small number of hops). Index-based methods, especially ones based on 2-hop labeling are often used for high query efficiency. However, the evolvements of small-world networks in many realistic scenarios pose the challenge of efficient maintenance of the shortest path index. In this work, we adopt the state-of-the-art Parallel Shortest-distance Labeling (PSL) as the underlying 2-hop labeling construction method, and design algorithms to support its efficient update given edge weight changes (increase and decrease). Specifically, we focus on weighted PSL (WPSL) and propose a propagation-based update mechanism for both synchronous and asynchronous propagation. We also identify the curse of pruning power in the edge weight increase case, and solve it with a balance between index size and effectiveness. Followed by, we extend the asynchronous propagation method to Pruned Landmark Labeling (PLL) for faster index maintenance and query processing with a smaller index size. Finally, we further optimize the index performance by reducing the index size through graph contraction and accelerating the index update through parallelized mix index update. Our experimental results on real-life and synthetic networks demonstrate the superiority of our algorithms over the relevant baselines on index maintenance.