SDaaS:一种传感流数据的服务化封装方法

发布时间：2018-03-02 11:40

本文选题：传感流数据　切入点：流数据服务　出处：《计算机学报》2017年02期 　论文类型：期刊论文

【摘要】：来自不同传感器网络的流数据共享和集成对于带动相关业务和行业的创新具有重要意义.现有的传感网络往往是任务导向或领域专用的,仅适用于特定的应用场景,难以有效地在不同应用间共享和重用其数据资源.传感流数据的服务化是一种有效解决物理传感网络数据资源共享和重用的方法.针对已有服务化方法在应对大规模传感流数据共享和用户并发访问方面存在的局限性,该文提出了一种面向传感流数据的服务化封装方法——SDaaS(Stream Data as a Service),该方法使用事件的方式驱动传感流数据的处理和传输,通过对传感数据的融合操作实现服务对传感流数据的深层次加工,并基于Pub/Sub机制实现传感流数据的按需分发.文中基于Spark Streaming实现对大规模流数据加工操作的封装,并通过对传统的基于匹配树的事件匹配算法进行改进实现了高效的流数据内容分发,以保障将传感数据实时的分发给相应需求.该文通过实验验证了流数据服务的性能,印证了流数据服务能够响应不同的数据需求,在毫秒级别将数据流分发给不同应用.
[Abstract]:Stream data sharing and integration from different sensor networks are important for driving innovation in related businesses and industries. Existing sensor networks are often task-oriented or domain-specific and can only be used in specific application scenarios. It is difficult to share and reuse the data resource among different applications effectively. The service of sensor flow data is an effective method to solve the problem of data resource sharing and reuse in physical sensor network. The limitations of scale sensor flow data sharing and concurrent user access, In this paper, a service-oriented encapsulation method for sensor stream data is proposed. The method uses event mode to drive the processing and transmission of sensor flow data. Through the fusion operation of sensor data, the deep processing of sensor flow data is realized by the service, and the on-demand distribution of sensor flow data is realized based on Pub/Sub mechanism. The encapsulation of large-scale flow data processing operation based on Spark Streaming is realized in this paper. By improving the traditional event matching algorithm based on matching tree, the efficient streaming data content distribution is realized to ensure the real-time distribution of sensor data to the corresponding requirements. The performance of streaming data service is verified by experiments. It is proved that streaming data services can respond to different data requirements and distribute data streams to different applications at millisecond level.
【作者单位】：天津大学计算机科学与技术学院;北方工业大学大规模流数据集成与分析技术北京市重点实验室;北方工业大学云计算研究中心;
【基金】：国家自然科学基金(61672042) 北方工业大学“人才强校计划”青年拔尖人才培育计划北京市市委组织部资助青年骨干个人项目(2015000020124G024)资助~~
【分类号】：TP393.09;TP301.6

【相似文献】