Optimized Data Loading for a Multi-Terabyte Sky Survey Repository

Abstract

Advanced instruments in a variety of scientific domains are collecting massive amounts of data that must be post-processed and organized to support scientific research activities. Astronomers have been pioneers in the use of databases to host highly structured repositories of sky survey data. As more powerful telescopes come online, the increased volume and complexity of the data collected poses enormous challenges to state-of-the-art database systems and data-loading techniques. When the data source is an instrument taking ongoing samples, the database loading must, at a minimum, keep up with the data-acquisition rate. These challenges are being faced not only by the astronomy community, but also by other scientific disciplines interested in building scalable databases to house multi-terabyte archives of complex structured data. In this paper we present SkyLoader, our novel framework for fast and scalable data loading that is being used to populate a multi-table, multi-terabyte database repository for the Palomar-Quest sky survey. Our framework consists of an efficient algorithm for bulk loading, an effective data structure to support data integrity and proper error handling during the loading process, support for optimized parallelism that matches the number of concurrent loaders with the database host capabilities, and guidelines for database and system tuning. Performance studies showing the positive effects of the adopted strategies are also presented. Our parallel bulk loading with array buffering technique has made fast population of a multi-terabyte repository a reality, reducing the loading time for a 40-gigabyte data set from more than 20 hours to less than 3 hours. We believe our framework offers a promising approach for loading other large and complex scientific databases.