The Application Of Parallel Computing To Spatial Scanning Of Seismicity Rate Changes (Z-Value) And Its Efficiency Evaluation

Seismic quiescence is a very important research aspect for seismic researchers. In particular, as the precursory seismic quiescence is proposed, the quantitative study on precursory seismic quiescence is rapidly motivated. Many precursory quiescence phenomena prior to medium and large earthquakes which validated the precursory seismic quiescence at large were reported. There are totally three methods widely used for studying seismic quiescence as follows: the spatial-temporal variation distribution of seismic activity, seismic rate changes and the systematical deviation of the earthquake releasing energy. Z-value is a method widely used to quantify and calibrate the seismic rate changes. There is a heavy computation problem to obtain good Z-value by taking high precision grid method and choosing the equal number of earthquakes as the spatial scanning window, and the spatial scanning of other seismic parameters have the similar mode of processing. Thus, it is very meaningful to find a solution to heavy computation problem.In the first part of paper, the research history of seismic quiescence and some quantitative computing methods at present are reviewed, and the problem resulting from using Z-test method for studying seismic rate changes is discussed, then research purpose and research contents are brought forth.In view of the long-term applied prospect of seismic data processing with parallel computing techniques, the second part of paper begins with the conception of parallel computing, and simply introduces the development background,structure models and access models of parallel computers, then emphatically discusses the mainstream development directions of parallel computing system architecture, processors family and parallel computing network from the point view of Top500 list.As a precondition of parallel computing, establishing a PC cluster is an important and key step. The third part of paper gives a step by step installation and configuration of cluster service softwares based on existing hardware condition, some of which include Network Information Service (NIS) for single user management, Network File System (NFS)for file sharing, Network Time Protocol (NTP) for uniform time, Torque/Maui for resources management and job scheduling, Ganglia for providing cluster runtime states information based on Web browser and MPICH for offering parallel computing environment. At the end of this part, cluster storage system is discussed with reference to the Storage Area Network (SAN) and Parallel Virtual File System (PVFS).The fourth part of paper is focus on parallel processing. For one thing, there is a detailed introduction of the method and process of Z-value spatial scanning parallel computing, for another, the rationality of Z-value parallel processing is validated by the comparison of the parallel computing maps with ZMAP computing maps in the same computing conditions.As parallel computing performance test is a hotspot in the parallel computing area, the fifth part of paper introduces the algorithm-level performance indexes at first, and makes analysis of Z-value parallel computing test results in the self-built PC cluster. The results show that the designed parallel program has very high speedup and parallel efficiency with different computing scales. In addition, the test results also indicate that the I/O operation time on a large scale of seismic data may take a considerable ratio in the total computing time. Finally, the scalability of the Z-value spatial scanning parallel program by equal average speed test is discussed at this part. As a result, it is indicated that the program has a good scalability, that is to say, the Z-value parallel program will be applied for a large-scale cluster with high performance.To meet the need for interactively obtaining the process results of seismic data in the research work, a method by CGI to realize parallel computing service has been used in the paper. Therefore, the sixth part starts with the conception and principle of CGI, and then explains the method and process of Z-value spatial scanning parallel computing service in detail, and shows some Z-value maps in the end.In conclusion, the last part of paper sums up some research results, and points out the shortcomings and some interesting aspects in the future research.