Parallel Computing Optimization for Processing the Seismic Data

Efficiency of explicitly parallel data processing programs is sometimes restricted by bottlenecks like file storage bandwidth or network capacity. An approach to the calculation time prediction based on few program execution parameter estimates is proposed. It can be used to select the best processor number for a specific computation, to optimize hardware architecture of a parallel computer for a set of typical tasks, to balance choice of hardware components for better real productivity under a budget limitation.