基于数据级任务分解的大范围地质灾害预警并行计算架构设计与应用

为实现大范围地质灾害预警实时分析计算,搭建了地质灾害预警并行计算架构。结合地质灾害隐患点的现状和监测设备的实际情况,以地质灾害隐患点为分析计算单元,采用数据级并行计算方式对地质灾害计算任务进行分解,实现了任务生成、分解、派发、计算等功能;采用MPI和Socket套接字相结合的方式,达到了并行计算与高效通信的目的;将该并行计算架构应用于某省2016年8月2~4日台风造成的大范围强降雨状况下的地质灾害预警计算实验验证。结果表明:在大范围地质灾害预警中,该并行计算架构随着执行进程数增加而获得一定的加速比,当执行进程达到16个时加速比可达5.652,明显缩短了传统串行计算所需要的时间,实现了地质灾害的实时预警。

The Design and Application of Parallel Computing Architecture for Large-scale Geological Disaster Warning Based on Data Level Task Decomposition

In order to realize the real-time analysis computing of large-scale geological disaster warning, the parallel computation structure of geological hazard early warning is set up. Taking hidden geological hazard points as analysis computing unit, combined with their current circumstances and the actual situation of monitoring equipment, data-level parallel computing method is adopted to decompose the geological disaster computing tasks, to complete task generation, decomposition, distribution, computation and so on. The efficient communication of parallel computation is realized through the combination of MPI and Socket sockets. The parallel computing framework is applied to a geological hazard early warning and calculation experiment in the case of large-scale heavy rainfall caused by the typhoon on August 2~4, 2016. The results show that the parallel computing framework has significantly shortened the time required for traditional serial computing and has implemented preliminarily the real-time warning of geological hazards in the early warning of large-scale geological disasters.