A Study On Data Placement And Retrieval Method Of Massive Multi-Band Astronomical Catalogs

An increasing number of telescopes are being built to provide multi-band data of celestial objects,which is indispensable to astronomical research.The amount of collected observation data,however,have put tremendous pressure on computing systems.Moreover,with the development of the observation equipment,the number of astronomical catalogs that telescopes generated per day keeps increasing rapidly.Multi-band observation data refers to the data about the same object obtained by telescopes in different band(e.g.optical telescope,radio telescope,etc.).Different types of features of a celestial object can be observed by telescopes of different bands.Astronomical observation image shows the images of the Crab Nebula observed in 6 different bands.Though celestial object target shows distinguishing characteristics in each band,the astronomical catalog is the standard file recording all the corresponding information no matter which band of observation it is.For astronomers,an efficient indexing service is needed to find related records among massive catalogs.However,the features of massive astronomical data complicate this goal.In this paper,DCSIS(Distributed Cone Search Indexing System),is proposed to solve the problems listed above.Major contributions of DCSIS include a new meta file format for astronomical catalogs,achieving scalability and parallelism for cone search,and the ability to flexibly add data to index system.Evaluations are performed on the Tianhe-1A supercomputer to showcase DCSIS'scalability for large scale deployment,the results of which show that DCSIS reduce the response time of Multi-band cone search into a tolerant range.Tianhe-1A supercomputer is used as the evaluation environment.The data of 2MASS,PPMXL,SDSS and other truthful astronomical catalogs is acquired from National Astronomical Observatories of the Chinese Academy of Sciences(NAOC)as the data set.Evaluation result shows that DCSIS have great scalability and shorter response time compared to other databases.