Back to Volume
Paper: The Indexing of the SDSS Science Archive
Volume: 216, Astronomical Data Analysis Software and Systems IX
Page: 141
Authors: Kunszt, P. Z.; Szalay, A. S.; Csabai, I.; Thakar, A. R.
Abstract: The Spatial Indexing used in the Sloan Digital Sky Survey (SDSS) Science Archive divides the spherical surface into triangles in a hierarchical scheme resulting in roughly equal surface areas at each level, which is a big advantage over other schemes. The location of a point on the sky may be given by the unique index id to any level, refining it with each step. This naming scheme is being used successfully in other catalogs, too, like GSC-II and GAIA. The use of the Spatial Index in the SDSS is two-fold, a level-5 index is used to partition the bulk data, and a high-resolution level-14 index id is assigned to each data point to enable quick lookup and proximity searches. Use of this indexing scheme in more catalogs will enormously simplify cross-matching of objects. Using a new computing paradigm, we recently realized a quantum leap in performance that makes this scheme competitive with bit-interleaving and requires very little memory. The Flux-space Indexing used is a traditional k-d tree. The space is 5 dimensional, 5 being the number of SDSS-filters. The specialization to astronomical data has been achieved by modeling the location of the main branch in this space and applying the k-d tree subdivisions only to its confined area. The outliers are indexed separately. Most of the interesting data points come directly from the outlier part of the index, with no additional analytical effort. Additionally, the key-lookup index feature of object-oriented databases is exploited for much-used parameters like flags.
Back to Volume