Design of Distributed Parallel Computing Using by MapReduce/MPI Technology

Abstract

Abstract. This paper describes a constructive approach of distributed parallel computing using by hybrid union of MAPREDUCE and MPI technologies for solving oil extracting problems. We extend a common architecture of MAPREDUCE model by organizing decomposition of computational domain at different stages of MAPREDUCE process. We describes Model Driven Architecture (MDA) models for developing formal views of high-performance computing technologies using MAPRE-DUCE. We made computing experiments and show on specific HPC infrastructure. All implementations of programs is realize on Java platform. This approach will possible one of the ways to do cloud computing on high performance heterogeneous systems.