BatchHL+: batch dynamic labelling for distance queries on large-scale networks

Abstract

Many real-world applications operate on dynamic graphs to perform important tasks. In this article, we study batch-dynamic algorithms that are capable of updating distance labelling efficiently in order to reflect the effects of rapid changes on such graphs. To explore the full pruning potentials, we first characterize the minimal set of vertices being affected by batch updates. Then, we reveal patterns of interactions among different updates (edge insertions and edge deletions) and leverage them to design pruning rules for reducing update search space. These interesting findings lead us to developing a new batch-dynamic method, called BatchHL+ , which can dynamize labelling for distance queries much more efficiently than existing work. We provide formal proofs for the correctness and minimality of BatchHL+ which are non-trivial and require a delicate analysis of patterns of interactions. Empirically, we have evaluated the performance of BatchHL+ on 15 real-world networks. The results show that BatchHL+ significantly outperforms the state-of-the-art methods with up to 3 orders of magnitude faster in reflecting updates of rapidly changing graphs for distance queries.