A Data Partitioning Method for Parallel Digital Terrain Analysis

Abstract

Parallel computing of the intensive data is one of the effective methods to improve high-performance computation of massive data. The purpose of this paper is to study the method of data partitioning and scheduling which is geared to the strategy of parallel computation facing to the distribution of data in sequence. According to the features of the intensive data computation, this paper puts forward the concept of data granularity, return granularity and saturation. The parallel computing and scheduling model facing to the distribution of data in sequence is given based on these concepts. Considering that the startup and shutdown of data distribution has some overhead, the distribution of a data block in sequence and the calculation of another data block can be done at the same time. While the total time of the calculation is not decreasing with the increase of the number of data blocks divided, there exists an optimal value. Through analyzing the slope algorithm of Digital Terrain Analysis (DTA), the optimal solution of data partitioning and the best number of computing nodes is presented in this paper. The results of the experiment show that the theoretical analysis and the results of the experiment are basically consistent.