| 梁彦* **,王春祥*,曹华伟* **,范东睿* **.基于自适应增量计算模型的流式图分析优化[J].高技术通讯(中文),2026,36(3):244~255 |
| 基于自适应增量计算模型的流式图分析优化 |
| Optimization of stream graph analysis based on adaptive incremental computation model |
| |
| DOI：10. 3772 / j. issn. 1002 - 0470. 2026. 03. 003 |
| 中文关键词: 流式图； 增量计算； 传统计算； 流式图处理 |
| 英文关键词: stream graph, incremental computing, traditional computing, streaming graph processing |
| 基金项目: |
| 作者 | 单位 | | 梁彦* ** | （*处理器芯片全国重点实验室(中国科学院计算技术研究所)北京 100190）
（**中国科学院大学北京 100049） | | 王春祥* | | | 曹华伟* ** | | | 范东睿* ** | |
|
| 摘要点击次数: 38 |
| 全文下载次数: 38 |
| 中文摘要: |
| 流式图在计算机网络、地图规划、社交网络等领域广泛应用。处理流式图时，实时计算至关重要，但传统方法需对整个数据集重新计算，效率低且资源消耗大。增量计算通过修正历史查询结果减少计算量，但随着迭代次数增加，其效率可能下降。本文提出一种创新的图计算优化方法，采用机器学习和数学建模评估不同计算模型的性能；提出基于迭代轮数的模型切换策略，结合传统与增量计算的优势，显著提升了计算效率。此外，针对内存开销问题，本文提出减少增量计算迭代次数的方法，以降低内存占用。优化后，系统性能比传统模型和增量计算模型分别提升1.39倍和1.14倍。 |
| 英文摘要: |
| Stream graphs are widely applied in fields such as computer networks, map planning, and social networks. Real-time computation is crucial for processing stream graphs, but traditional methods require recomputing the entire dataset, leading to low efficiency and high resource consumption. Incremental computation reduces the computational load by correcting historical query results; however, as the number of iterations increases, its efficiency may decline. This paper proposes an innovative optimization method for graph computation, employing machine learning and mathematical modeling to evaluate the performance of different computation models. It introduces a model switching strategy based on the number of iterations, effectively combining the advantages of traditional and incremental computation to significantly enhance computational efficiency. Additionally, to address memory overhead issues, this study proposes a method to reduce the number of iterations in incremental computation, thereby lowering memory usage. After optimization, the system achieved performance improvements of 1.39 times and 1.14 times compared to traditional and incremental computation models, respectively. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
