An Efficient Method For Indexing And Retrieving Astronomy Images Based On Pixelization Of Sphere

Astronomy has always been,and will continue to be a data-based science,and FITS images captured by telescopes make up a significant part of all the available astronomy data.With the number of both telescopes and massive sky survey projects on constant increase,the total size of astronomy data is growing rapidly.Currently,astronomy data is on terabyte size and speeding towards petabytes.It's safe to say that the science of astronomy has entered the Big Data era.The overwhelming amount of data challenges the traditional procedures and tools for astronomy data analysis.Usually,an astronomer has a specific target in mind and do not need nor is able to analyze the entirety of the available data.Therefore,the first and foremost step towards efficiently utilizing Astronomy Big Data is to efficiently locate the desired subset of data.Our work presented in this paper is based on the actual need of astronomers and aims to provide an efficient method to query an enormous amount of FITS image files.Our method,namely AQUAdex,follows the in-situ data analysis model,which requires no preloading since it can index and query directly on the original FITS files.The index is kept at a small size by only taking essential information into the index so that the entire index data can fit in the RAM of servers.In this way,AQUAdex is able to perform queries all in RAM and saves a lot of time-consuming disk I/Os.Furthermore,AQUAdex is also scalable and is fit for the multi-core multi-node server structure.AQUAdex is used to build index on real observation data,and experimental results has demonstrated that its index structure is over 9 times more efficient than that without a pseudo-sphere index.AQUAdexIM(short for AQUAdex In Memory)is also 10 times more efficient than Redis,a popular In Memory Database.Its scalability is attested in the multi-core multi-node experiments.