Exploiting Scalable Parallelism for Remote Sensing Analysis Models by Data Transformation Graph

摘要

According to the great hunger in performance capability and scalability for remote sensing analysis models, it is important to exploit scalable parallelism for remote sensing data analysis models. In this paper, a method named data transformation graph (shortly DTG) is introduced, which describes an analysis model by transformations among data items. DTG can be used to study the solvability and performance of analysis models. Taking global drought detection as an example, its execution and optimization are studied carefully by DTG, and some methods are proposed for accelerating remote sensing data analysis models. At last, a distributed data-intensive computing test system is built based on Robinia, and global drought detection application is implemented for performance evaluation. The test result shows that DTG based paral-lelization and optimization improves the performance with high efficiency evidently, and DTG is valuable to study and optimize remote sensing data analysis models for higher performance in distributed and parallel computing environments.