面向海量数据的并行天文交叉证认  

Parallel massive data oriented astronomical cross-match

作　　者：赵青[1] 孙济洲[1] 于策[1] 崔辰州[2] 肖健[1] 

机构地区：[1]天津大学计算机科学与技术学院,天津300072 [2]中国科学院国家天文台,北京100012

出　　处：《计算机应用》2010年第8期2056-2059,2206,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(10978016);天津自然科学基金资助项目(08JCZDJC19700);天津市科技支撑重点项目(09ZCKFGX00400)

摘　　要：交叉证认是实现多波段数据融合的关键技术,天文数据的海量性使这一问题必须要依靠计算机技术加以解决。按照PCAM并行设计模型设计了并行交叉证认算法。针对交叉证认在数据I/O访问方面存在的性能瓶颈,通过调整划分的粒度、过滤空白区域、优化数据加载、计算流程等方法,协调了数据读取量与计算量间的关系。实验表明该并行方法对交叉证认计算的效率提升明显。另一方面还考虑了对HTM、HEALPix两种最常用天文数据索引方式的支持,并通过实验对比了两者的性能,为我国天文数据主题库、虚拟天文台等项目提供了技术参考。Cross-match is the kernel technology to realize multi-band data aggregation.It must be resolved by computer technologies since the astronomical data sets are very large.This paper issues a parallel cross-match function using PCAM parallel programming model.Because the performance bottle-neck of cross-match exists on data I/O accessing this function balanced the amount of data reading and computation by tuning partition granularity removing blank areas improving data load and computation flow and etc.Experiments prove that it can speed up large-scale cross-match greatly.On the other hand the support of both of the two most famous astronomical data index functions HTM and HEALPix were considered and the performance comparison between them were analyzed through experiments.The high-efficiency cross-match function in this paper offers a technology conference for the projects such as national astronomical database and virtual observatory.

关 键 词：天文交叉证认 HTM HEALPix PCAM并行程序设计模型 大规模数据处理 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TP302[自动化与计算机技术—控制科学与工程]