采用MapReduce模型的甚长基线干涉测量并行处理方法

发布时间：2018-04-11 14:09

本文选题：甚长基线干涉测量 + 数据相关处理　；参考：《西安交通大学学报》2015年04期

【摘要】：针对传统甚长基线干涉测量(very long baseline interferometry,VLBI)并行处理方法存在逻辑复杂、数据存储和计算扩展性较差等缺点,基于MapReduce模型设计了一种VLBI观测数据相关的并行处理方法。该方法采用模型周期作为数据并行处理基本单元,构建数据关联算法分割观测数据使各台站同一模型周期数据对齐到同一波前,设计实现数据相关计算的Map过程,使用Reduce对各模型周期结果进行综合处理以得到最终结果。实验结果表明:该方法在积分周期包含的模型周期数量大于节点CPU核数时性能比传统方式低25%左右,在积分周期包含的模型周期数量小于节点CPU核数时通常能提供更高的计算效率,并且能简化VLBI并行相关处理的复杂度,具有更好的数据存储和计算可扩展性。
[Abstract]:Aiming at the shortcomings of the traditional parallel processing method of very long baseline interferometry long baseline interferometry (VLBI), such as complicated logic and poor scalability of data storage and computation, a parallel processing method related to VLBI observation data is designed based on MapReduce model.In this method, the model period is used as the basic unit of data parallel processing, and the data association algorithm is constructed to divide the observation data so that the data of the same model period are aligned to the same wave front at each station, and the Map process of data correlation calculation is designed.Reduce is used to synthesize the periodic results of each model to obtain the final results.The experimental results show that the performance of this method is about 25% lower than that of the conventional method when the number of model cycles included in the integral period is larger than the number of node CPU kernels.When the number of model cycles included in the integral period is less than the number of node CPU kernels, it can provide higher computational efficiency, simplify the complexity of VLBI parallel correlation processing, and have better data storage and computational scalability.
【作者单位】：西安交通大学电子与信息工程学院;西安卫星测控中心;
【分类号】：TP338.6

【参考文献】