Toward Ubiquitous MapReduce Processing
Toward Ubiquitous MapReduce Processing
复制标题
迈向无处不在的 MapReduce 处理
DOI：
10.1109/iucc-css.2016.010
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
I. Satoh
中科院分区：
文献类型：
--
作者：
I. Satoh
无处不在的计算环境在网络的边缘产生大量数据，例如测量或监控真实世界的传感器。要分析这些数据，我们需要将数据传输到高性能服务器集群，例如数据中心和云计算。然而，数据传输的成本很高。为了使无处不在的计算环境能够分析数据，我们建议使用MapReduce，这是一种在为高性能服务器设计的大数据处理中分析数据的流行方法。本文提出的框架将数据处理程序部署在包含目标数据的节点上，作为映射步骤，并使用本地数据执行程序。最后，它将程序的结果聚合到特定节点作为约简步骤。文中描述了该框架的体系结构、基本性能和应用。
Ubiquitous computing environments generate a large amount of data at the edges of networks, e.g., sensors, which measure or monitor the real world. To analyze such data, we needed to transmit the data to a cluster of high-performance servers, e.g., data centers and cloud computing. However, the cost of data transmission is heavy. With the aim of enabling ubiquitous computing environments to analyze the data, we propose using MapReduce, which is a popular approach to analyzing data in big data processing that have been designed for high-performance servers. The framework proposed in this paper deploys programs for data processing at the nodes that contain the target data as a map step and executes the programs with the local data. Finally, it aggregates the results of the programs to certain nodes as a reduction step. The architecture of the framework, its basic performance, and its application are also described here.