期刊文献+

基于结构分解的分布式RDF图模式匹配优化算法

PATTERN MATCHING OPTIMIZATION ALGORITHM FOR DISTRIBUTED RDF GRAPHS BASED ON STRUCTURE DECOMPOSITION
下载PDF
导出
摘要 在分布式RDF查询处理中,由于查询图规模的不断增长,其复杂的结构使得查询优化面临着查询准确性和性能的双重挑战。针对这一问题,提出一种由结构主导的分布式子图匹配算法SDSM。根据查询图各部分结构的匹配特征,提出查询图的CPM分解模型;结合基于类型的摘要统计,通过代价模型得到查询起始节点及核心结构权重图,权重图与最小生成树思想相结合得到最优的查询匹配序列;各计算节点分两阶段进行查询,并在主机上进行轻量级的连接。实验将SDSM与StarMR、TriAD、Wukong等算法进行对比。结果表明,在处理复杂查询时,SDSM具有更高的查询效率,在多台机器以及不同数据集规模上的实验表明,SDSM也具有良好的扩展性。 In distributed RDF-based query processing,the complex structure of query graph results into the challenges of query optimizer in terms of query accuracy and query performance as eruptive growth of query graph scale.In order to solve this problem,a distributed subgraph matching algorithm SDSM was proposed on the basis of query graph structure.A CPM decomposition model was proposed according to the matching characteristics of each substructure in a query graph.A query starting node and a weight graph of core substructure were obtained through a cost model with the type-based summary statistics,then a optimize query matching sequence was obtained by the combination of weight graph and minimum spanning tree.The query was carried out in two stages in each computing node,then a lightweight connection was carried out on the host.The experiment compared SDSM with StarMR,TriAD,Wukong.The results show that SDSM has higher time-efficiency of complex queries and it has a better scalability on multiple machines and different dataset scales.
作者 孙云浩 邢维康 李冠宇 韩冰 李逢雨 Sun Yunhao;Xing Weikang;Li Guanyu;Han Bing;Li Fengyu(College of Information Science and Technology,Dalian Maritime University,Dalian 116026,Liaoning,China)
出处 《计算机应用与软件》 北大核心 2022年第6期228-238,共11页 Computer Applications and Software
基金 国家自然科学基金项目(61976032,61602076,61702072) 辽宁省自然科学基金项目(20170540144,201705400231,20180540003)。
关键词 分布式 RDF图模式匹配 结构分解 以类型为中心的评估 查询处理 Distribution RDF graph pattern matching Structure decomposition Type-centric evaluation Query processing
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部