Gopher: Efficient Dynamic Graph Pattern Mining via DAG-Driven Execution

Authors

• Yi Zhang, Huazhong University of Science and Technology
• Yu Huang, Huazhong University of Science and Technology
• Chaoqiang Liu, Huazhong University of Science and Technology
• Haifeng Liu, Huazhong University of Science and Technology
• Juntao Chen, Huazhong University of Science and Technology
• Jingrui Yuan, Huazhong University of Science and Technology
• Jianhui Yue, Michigan Technological UniversityFollow
• Xiaofei Liao, Huazhong University of Science and Technology
• Hai Jin, Huazhong University of Science and Technology
• Jingling Xue, UNSW Sydney

Document Type

Conference Proceeding

Publication Date

4-26-2026

Department

Department of Computer Science

Abstract

Graph pattern mining is essential for analyzing dynamic networks, where graphs evolve over time. To accommodate these changes, existing solutions update match sets incrementally, avoiding the need to re-mine the entire graph and achieving significant performance improvements. However, these methods suffer from inefficiencies due to redundant set intersection operations across subgraph instances, causing performance degradation. In this paper, we propose Gopher, a DAG-driven dynamic graph pattern mining system that leverages computation locality for enhanced performance. Gopher uses DAGs to represent set operations, identifying and merging common subexpressions at compile time. This reduces redundant computations during runtime. To maximize performance, we introduce a DAG-based mining engine with fine-grained parallelism to decouple data dependencies and a canonical expression-based restoration module to efficiently assemble the results. Evaluations on a multi-core CPU show that Gopher outperforms state-of-the-art solutions, achieving speedups of up to 75.58×, 35.99×, and 11.76× over Tesseract, Cheetah, and PSMiner, respectively.

Publisher's Statement

© 2026 Copyright held by the owner/author(s). Publisher’s version of record: https://doi.org/10.1145/3767295.3769365

Publication Title

Eurosys 2026 Proceedings of the 2026 European Conference on Computer Systems

ISBN

9798400722127

Recommended Citation

Zhang, Y., Huang, Y., Liu, C., Liu, H., Chen, J., Yuan, J., Yue, J., Liao, X., Jin, H., & Xue, J. (2026). Gopher: Efficient Dynamic Graph Pattern Mining via DAG-Driven Execution. Eurosys 2026 Proceedings of the 2026 European Conference on Computer Systems, 1722-1737. http://doi.org/10.1145/3767295.3769365
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p2/2645

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Version

Publisher's PDF