SkyQuery: An Implementation of a Parallel Probabilistic Join Engine for Cross-Identification of Multiple Astronomical Databases

TL;DR

SkyQuery introduces a parallel, probabilistic engine for efficient cross-identification of celestial objects across multiple large astronomical catalogs, leveraging SQL extensions and distributed computing.

Contribution

The paper presents a novel parallel system that performs Bayesian probabilistic cross-identification of multiple astronomical catalogs using extended SQL queries on commodity server clusters.

Findings

Handles billions of objects efficiently

Uses Bayesian probabilistic algorithms for accurate identification

Employs parallel processing with SQL extensions

Abstract

Multi-wavelength astronomical studies require cross-identification of detections of the same celestial objects in multiple catalogs based on spherical coordinates and other properties. Because of the large data volumes and spherical geometry, the symmetric N-way association of astronomical detections is a computationally intensive problem, even when sophisticated indexing schemes are used to exclude obviously false candidates. Legacy astronomical catalogs already contain detections of more than a hundred million objects while the ongoing and future surveys will produce catalogs of billions of objects with multiple detections of each at different times. The varying statistical error of position measurements, moving and extended objects, and other physical properties make it necessary to perform the cross-identification using a mathematically correct, proper Bayesian probabilistic algorithm, capable of including various priors. One time, pair-wise cross-identification of these large catalogs is not sufficient for many astronomical scenarios. Consequently, a novel system is necessary that can cross-identify multiple catalogs on-demand, efficiently and reliably. In this paper, we present our solution based on a cluster of commodity servers and ordinary relational databases. The cross-identification problems are formulated in a language based on SQL, but extended with special clauses. These special queries are partitioned spatially by coordinate ranges and compiled into a complex workflow of ordinary SQL queries. Workflows are then executed in a parallel framework using a cluster of servers hosting identical mirrors of the same data sets.