Research On Application Of Parallel Computing To Marine Water Age Spectrum Simulation

By studying water age and water age spectrum, marine environmental scientistscan understand the potential impact of the changes in water transport processes,therefore it has important theoretical and practical significance for ensuring thesustainable development of the coastal areas. In the current study of water age spectrumnumerical simulation, the computing efficiency is unsatisfactory becasue of the serialmodel used as well as the large amount of data and calculation involved. Especially inthe large-scale numerical simulation of the water age spectrum, the low efficiency ofcomputation will affect the research progress. The application of parallel computingtechnology in the water age spectrum numerical simulation can improve simulationspeed, so as to speed up the progress of research, it also has an important role in theapplication and development of parallel computing technology in the field of numericalsimulation of marine. This study was supported by the National Natural ScienceFoundation “Bohai multi-source water age and water age spectrum numerical studies(41176007)”.In this study, a hybrid parallel computing model of marine water age spectrum isestablished, based on the combination of Message Passing Interface (MPI) and OpenMulti-Processing(OpenMP) which are, respectively, suitable for distributed storageenvironment and shared storage environment. The hybrid model was used to simulatewater age spectrum of Bohai sea for test. The main work of this paper is as follows:(1) Established a hybrid parallel computing model of marine water age spectrum.On the basis of the presentation and analysis of the water age spectrum numericalsimulation process, combined two parallel technology MPI and OpenMP, realizedhybrid parallel computing of water age spectrum. Compared with the model of a singleparallel technology, the hybrid model is more suitable for multi-core cluster computersystem.(2) The water age of Bohai sea was simulated with the hybrid model and comparedthe results with serial model to verify the reliability. On this basis, the computingefficiency of the hybrid model was tested and compared with traditional pure MPImethod.(3) To learn the scalability of the hybrid mode further, the system memory resource usage was studied.The results show that, the hybrid MPI and OpenMP parallel model cansignificantly improve the computational efficiency of water age simulation, and featureswith higher computational efficiency and lower system resource consumption comparedto traditional pure MPI method.