通用遥感数据并行处理框架设计与实现

为了简化对地观测地面系统遥感数据并行处理软件的开发工作,在分析遥感数据处理流程、并行任务调度和容错策略的基础上,设计了遥感数据并行处理框架。该框架集成了遥感数据预处理并行任务调度、消息和数据交换、故障管理等公共功能,并设计实现了简单易用的插件接口规范,以支持多载荷、多种遥感数据预处理功能的扩展以实现并行处理,具有较好的通用性和功能扩展性。最后基于消息传递接口(MPI)技术,在集群上实现了遥感数据处理并行框架,并在此框架上实现遥感数据预处理的并行系统,完成了系统并行性能测试与分析,结果表明该框架在简化遥感数据预处理并行功能开发的同时还能满足遥感数据并行处理效率要求。

Design and Implementation of Universal Parallel Framework for Remote Sensing Process

In order to simplify the development of remote sensing data parallel processing software on ground station, the universal parallel framework for remote sensing data process is designed based on analysis of remote sensing data process flow, job scheduling strategy and node fault-tolerance strategy. It integrates job parallel scheduling, message and data transferring and fault management and implements easily-using plug-in interfaces to support the extensibility of data process of multi-sensor and multi-functions. Finally, based on Message Passing Interface (MPI) technology, the framework is implemented on cluster. The parallel system for remote sensing data preprocess is implemented based on the framework as well. After the test and analysis of system performances, the results show that the framework can simply the development of data parallel process and meets the requirements of parallel computing efficiency.