NewSP: A New Search Process for Continuous Subgraph Matching over Dynamic Graphs

Ziming Li,Youhuan Li,Xinhuan Chen,Lei Zou,Yang Li,Xiaofeng Yang,Hongbo Jiang

DOI: https://doi.org/10.1109/icde60146.2024.00257

2024-01-01

Abstract:In this study, we address the problem of unnecessary computations in traditional continuous subgraph matching (CSM) frameworks due to premature expansions of the search space in dynamic graphs. Traditional CSM frameworks expand small partial matches according to a specific matching order until the final results are obtained. This extension involves two sequential steps: computing candidate vertices for an unmapped query vertex and expanding the search space using these candidate data. However, this long-established search model has a potential flaw, as premature expansions of the search space can lead to unnecessary computations. To address this issue, we introduce a novel search process, NewSP. Unlike traditional methods, NewSP emphasizes operations rather than extensions, incorporating a unique feature of postponing expansion at the operation level. This approach prevents premature expansions without compromising the initial pruning power of the selected matching order. Furthermore, NewSP allows for multiple consecutive expansions, paving the way for a multi-expansion strategy for further optimization. Our model also enables the implementation of cache strategies for candidate set reuse, as it does not necessitate immediate expansion of a candidate set once identified. To improve performance, we propose an adaptive index filtering strategy independent of the specific index used. Comprehensive experiments demonstrate that our method improves by up to two to three orders of magnitude compared to traditional algorithms. A case study showed that NewSP can accelerate the majority of subgraph matching algorithms.