近海生态环境预评估系统并行化与可视化研究

发布时间：2018-05-02 18:46

本文选题：海洋数值模式 + 并行计算　；参考：《河北工程大学》2016年硕士论文

【摘要】：海洋数值模式有着庞大的计算区域以及复杂的求解过程,因此具有浩大的计算规模,模式最终计算结果也包含了海量数据。采用并行计算方能有效提升计算效率并在预期时间内完成计算任务;应用并行技术的同时对并行程序的优化也是提高应用整体性能不可或缺的步骤,其中I/O优化便是一个个关键所在。在完成应用效率优化的基础上,使用数据可视化与Mashup技术将隐藏于数据中的现象或关联以及数值数据与现实地理信息之间的关系可视化,可以有效帮助分析理解数据和寻找数据中的规律。本文使用基于MPI的并行技术对我国近海生态环境预评估系统中,胶州湾示范区的物理-生态耦合模式进行并行化处理,然后对并行程序进行了进一步优化;完成应用性能提升后针对其计算结果进行了数据可视化处理并与百度地图进行了混搭,主要工作如下:对原串行模式进行并行化处理并对效能的提升进行测试与评估;主要实现了将基于MPI的并行计算技术应用到胶州湾示范区的物理-生态耦合模式中,其基本思想是通过对整个计算区域的划分,使每个子区域由不同计算节点并发完成计算,从而提高系统计算效率。虽然模式中计算模块已经进行了并行化处理,I/O模块仍采用传统的串行I/O方式,在一定程度上降低了应用性能。因此使用基于MPI-2规范提出的并行I/O算法对I/O模块进行并行化处理,并与传统几种串行I/O进行对比分析。在完成应用性能提升后,采用基于HTML5的HeatMap技术实现模式计算结果的可视化,通过Mashup技术将模式计算数据与现实地理信息进行服务组合,通过AJAX引擎实现交互式控制数据的抽取和画面的显示,使得隐藏于数据中的现象和关联以及数据与现实之间的关系变得可视化,为分析理解数据和寻找数据中的规律提供了一种强有力的手段。实验结果表明通过采用并行计算技术以及并行I/O技术对胶州湾示范区的物理-生态耦合模式进行并行化,大大提升了运行速率;采用基于Mashup的可视化技术为最终数据中有效信息的挖掘、科研人员之间的交流以及更好的将科研数据服务于决策者和普通用户提供了有力帮助。
[Abstract]:The ocean numerical model has a large computational area and complex solution process, so it has a large scale of calculation, and the final results of the model also contain a large amount of data. Parallel computing can effectively improve the computing efficiency and complete the computing task in the expected time, and the application of parallel technology to optimize the parallel program is also an indispensable step to improve the overall performance of the application. I / O optimization is the key one by one. Based on the optimization of application efficiency, the phenomenon or association hidden in data and the relationship between numerical data and real geographic information are visualized by using data visualization and Mashup technology. It can help to analyze and understand the data and find the rules in the data. In this paper, the parallel technology based on MPI is used to parallel the physical-ecological coupling model in Jiaozhou Bay demonstration area, and the parallel program is further optimized. After the application performance is improved, the data visualization is processed and mixed with Baidu map. The main work is as follows: parallelizing the original serial mode and testing and evaluating the efficiency; The parallel computing technology based on MPI is mainly applied to the physical-ecological coupling model in Jiaozhou Bay demonstration area. The basic idea is to divide the whole computing area into different computing nodes to accomplish the calculation simultaneously. In order to improve the efficiency of the system calculation. Although the computing module in the mode has been parallelized, the I / O module still adopts the traditional serial I / O mode, which reduces the performance of the application to a certain extent. Therefore, parallel I / O algorithm based on MPI-2 specification is used to parallelize the I / O module, and the parallel I / O algorithm is compared with the traditional serial I / O algorithm. After the application performance is improved, HeatMap technology based on HTML5 is used to realize the visualization of pattern calculation results, and the data of pattern calculation and real geographic information are combined with Mashup technology. The interactive control data extraction and picture display are realized by AJAX engine, which makes the phenomena and associations hidden in the data and the relationship between data and reality become visual. It provides a powerful means to analyze and understand the data and find the rules in the data. The experimental results show that the parallel computing technology and parallel I / O technology are used to parallelize the physical-ecological coupling model in Jiaozhou Bay demonstration area, which greatly improves the running rate. The visualization technology based on Mashup provides powerful help for the effective information mining in the final data, the communication among researchers and the better service of scientific data to decision makers and ordinary users.
【学位授予单位】：河北工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：X826

【参考文献】