The Performance Comparison and Evaluation of Parallel Platforms based on Throughput Model

摘要

Economical and high-efficient parallel computing platform are of great significance to remote sensing image processing. The throughput ability of every step of such processing heavily affects the application of remote sensing image parallel processing technologies. And throughput is a key performance index of high price/performance rate computing platform. Through analyzing the throughput model, we can obtain the characteristics of the parallel processing system throughput, also quantitatively analyze the throughput ability of each processing step brought by various peripheries and their matching relationships. This paper discusses the throughput characteristics of super computer and common COW (cluster of workstation), which are typical parallel computing platforms. And the study of real cases has been given to show the quantitative analysis method to construct economical and high-performance computing platform according to throughput characteristics