Research On Seismic Ray-tracing Method Based On MapReduce

While the pre-stack Kirchhoff integral depth migration obtained a series successin solving complex structural imaging since the late 1980s, the ray tracing algorithm,as one of the foundations, had also been a lot of promotion and development, and a lotof new algorithms which were different from the traditional methods. Ray tracingmethod as a rapid and effective field approximation calculation, not only of greatsignificance for seismic wave theory, but also be directly applied to the seismic waveinversion and migration imaging process.Cloud computing was developed based on the distributed computing, parallelcomputing and grid computing after the rapid development of Internet broadbandtechnology and virtualization technology. The predecessor of cloud computing was akind of grid computing used to solve large problems by parallel computing , andprovided utility computing which treated computing resources as a quantifiableservice. Cloud computing can implement distributed storage of data resources in thecalculation, the calculation process parallelism, and thus can take full advantage ofexisting computing conditions to explore the potential of cloud computing. Therefore,it would be a great significance to port existing algorithms to Hadoop platform so thatthey can execute a large amount of data computing tasks parallel.The main work in this paper were in-depth study of cloud computing,MapReduce technique and the basic theory, models and algorithms of ray tracing ,made the analysis and improvement of shortest path ray tracing algorithm, travel-timelinear interpolation algorithm and travel-time parabolic interpolation algorithm. thealgorithms were improved parallel by MapReduce technology, and the improvementideas and implement process were proposed in the paper. At last, we evaluated theimproved algorithm and experiment showed that the alg had good performance whendata has large scale. the detailed works were as following:1. Did the research on the basic theory and the principles of parallel computing,cloud computing and MapReduce technologyFirstly, we researched the mechanism of Message Passing Interface(MPI)oftraditional parallel program and discussed the framework of OpenMP, whichimplemented the MPI in detail. Secondly, Cloud computing was developed based onparallel computing and based on parallel computing, we discussion the concept ofcloud computing, and classified the technology of cloud computing. At last, we didthe detailed analysis and discussion about MapReduce, including the programmingmodel, execute process and the principles of MapReduce. We also researched in detailabout the job's submission, initiation, distribution, and execution.2. Did the research about the basic theory and principles of ray-tracing, anddiscussed the parallelization about the algorithmwe summarized the two principles of one law about the ray-tracing, researchedthe basic process of ray-tracing. we also research the derivation process of the formulain the algorithms, including travel-time linear interpolation algorithm and parabolicinterpolation algorithm. based on above,we analyzed the complexity of orignalgorithm by time and space, and then the improvement of parallelization about the algorithm was design. we gave the idea and analyzed the complexity of the improvedalgorithm.3. Designed and implemented the ray-tracing algorithm based on Hadoop andMapReduce and tested itWe researched the process of building the Hadoop platform and developingenvironment and implemented the parallelized shortest path ray-tracing method andtravel-time linear interpolation algorithm and tested them. The experiment showedthat when processing the large scale data, the Hadoop-based algorithm had goodperformance on the efficiency.