Evaluating MapReduce for seismic data processing using a practical application

Abstract

Huge amounts of seismic data undergo complex iterative processing in the oil industry to get knowledge of the earth’s subsurface structure to detect where oil can be found and recovered.To evaluate the suitability of MapReduce for seismic processing algorithms,the algorithm design and implementation of Fresnel tomography on Hadoop MapReduce was described.Experiments demonstrate that MapReduce is approximately 3 times slower than MPI,and tuning the performance of MapReduce is really hard.To expand its applicability to high performance computing for oil industry,MapReduce should be improved in the flexibility and provide the opportunity to exploit fine-grained thread-level parallelism.Finally,research ideas to achieve these objectives were presented.

Keywords

• oil exploration;MapReduce;Fresnel tomography;high performance computing;seismic data