The Study And Implementation Of Molecular Dynamics Simulation Based On Hadoop

Hadoop is a more popular cloud computing tool. In solving the problem of huge amountsof data, compared to expensive and excellent mainframe computers, Hadoop has very obviousadvantages. Meanwhile, Molecular dynamics simulation is a powerful method to study complexmatter. Since the MD’s data is huge, MD has increasingly performance requirements tocomputer performance. Running molecular dynamics simulation programs on Hadoop hasmany research meanings such as saving hardware and software investment and shorteningsimulation time.Based on atom-decomposition method and MapReduce programming, we introduce theinteraction force between molecules need the longest calculate time implementation on Hadoop.The experiment results show molecular dynamics simulation on Hadoop obtained good resultsas high as20x speedup.According to the thermodynamic quantities can parallel computing intermolecular potentialfunction, we used the Hadoop to design and implement parallel algorithm. The results indicatethat the Hadoop achieves very high speed up of up to near170folds compared withsingle-based algorithm.When calculating the radial distribution function, each molecule using adjacent moleculesaround search algorithm to find the number of molecules can be parallel. So we used theHadoop to implement parallel algorithm. The results show that it can improve300timesspeedup than ordinary single algorithm.