A United Framework for Large-Scale Resource Description Framework Stream Processing
A United Framework for Large-Scale Resource Description Framework Stream Processing
复制标题
DOI：
10.1007/s11390-019-1941-9
复制
发表时间：
2019-07
影响因子：
0.7
通讯作者：
Hong Fang;Bo Zhao;Xiaowang Zhang;Xuanxing Yang
中科院分区：
文献类型：
--
作者：
Hong Fang;Bo Zhao;Xiaowang Zhang;Xuanxing Yang
资源描述框架（RDF）流有助于对时空数据进行建模。在本文中，我们提出了一个用于大规模RDF流处理的框架，LRSP，用于处理大规模RDF流上的一般连续查询。首先，我们提出一种形式化（命名为CT-SPARQL），以统一的、明确的方式表示一般的连续查询。其次，基于我们的形式化，我们提出LRSP通过分离RDF流处理、查询解析和查询执行，以一种通用的白盒方式处理连续查询。最后，我们在一些基准数据集和实际数据集上，利用这些流行的连续查询引擎实现和评估LRSP。由于LRSP的体系结构，可以直接使用RDF的许多高效查询引擎（包括集中式和分布式引擎）来处理连续查询。实验结果表明，LRSP在处理大规模真实数据方面具有较高的性能。
Resource description framework (RDF) stream is useful to model spatio-temporal data. In this paper, we propose a framework for large-scale RDF stream processing, LRSP, to process general continuous queries over large-scale RDF streams. Firstly, we propose a formalization (named CT-SPARQL) to represent the general continuous queries in a unified, unambiguous way. Secondly, based on our formalization we propose LRSP to process continuous queries in a common white-box way by separating RDF stream processing, query parsing, and query execution. Finally, we implement and evaluate LRSP with those popular continuous query engines on some benchmark datasets and real-world datasets. Due to the architecture of LRSP, many efficient query engines (including centralized and distributed engines) for RDF can be directly employed to process continuous queries. The experimental results show that LRSP has a higher performance, specially, in processing large-scale real-world data.