An efficient parallel computing strategy for the processing of large GNSS network datasets
An efficient parallel computing strategy for the processing of large GNSS network datasets
复制标题
处理大型 GNSS 网络数据集的高效并行计算策略
DOI：
10.1007/s10291-020-01069-9
复制
发表时间：
2021
期刊：
影响因子：
4.9
通讯作者：
Lu Zhiping
中科院分区：
文献类型：
--
作者：
Cui Yang;Chen Zhengsheng;Li Linyang;Zhang Qinghua;Luo Sheng;Lu Zhiping
全球导航卫星系统（GNSS）已成为大地测量和地球动力学研究不可或缺的工具，并在过去几年中迅速发展，拥有丰富的地面网络、现代化的星座和多种信号频率。然而，由于站和卫星数量不断增加，数据处理负担显着增加。在本文中，提出了一种改进的并行计算方法来处理大型 GNSS 网络数据集。首先，通过分析GNSS网络数据并行综合处理的实用性，提出了针对传统GNSS处理模型的并行化策略。此外，为了最大限度地发挥现代微处理器和局域网环境的优势，我们提出了传统GNSS数据处理方法的多核并行计算；然后，该产品作为面向服务的架构发布，可以通过互联网中的多个节点找到和调用。显然，该方法结合了多核并行和网络并行的优点。实验表明，随着GNSS数据的积累和节点的增加，该方法的效率可以进一步提高。例如，在2000个站点的网络中，4个四核节点的并行方案的效率比传统串行方案至少快8倍。所有结果表明，所提出的策略是处理大型 GNSS 网络数据集的有效且有前途的方法。
The Global Navigation Satellite System (GNSS) has been an indispensable tool for geodetic surveying and geodynamics research and has rapidly developed over the past few years with abundant ground networks, modern constellations and multiple signal frequencies. However, due to increasing numbers of stations and satellites, the data processing burden has increased significantly. In this contribution, an improved parallel computing method is proposed for processing large GNSS network datasets. First, a parallelization strategy is introduced for the traditional GNSS processing model by analyzing the practicability of parallel integrated processing for GNSS network data. In addition, to maximize the advantages of modern microprocessors and local area network environments, we present the multi-core parallel computing of traditional GNSS data processing methods; then, the product is released as a service-oriented architecture that can be found and invoked through multiple nodes in the Internet. Obviously, this method combines the advantages of multi-core parallelism and network parallelism. Experiments show that the efficiency of the proposed method can be further increased with the accumulation of GNSS data and additional nodes. For example, in a network with 2000 stations, the efficiency of the parallel scheme with four quad-core nodes is at least 8 times faster than that of the traditional serial scheme. All the results demonstrate that the proposed strategy is an efficient and promising approach for processing large GNSS network datasets.